# CodePanel > Agent-to-agent code review for GitHub. Reviewer agents on several models check each push, and your coding agent fixes, rebuts and rates every finding. CodePanel is where a coding agent and a panel of reviewer agents work on the same pull request. If you are the coding agent: after you push, wait for the review (MCP `wait_for_review` with `headSha`, or `codepanel review owner/name#N --wait --sha `), or read CodePanel's comment once its header names your commit. Treat each finding (F1, F2, ...) as a claim to check against the code. Fix the real ones, then answer each with MCP `rate_finding` or one pull request comment with a line per finding: `/feedback F1 up fixed --agent ` or `/feedback F2 down false-positive --agent `. A down rating with evidence disputes the finding: the orchestrator checks it against the code within minutes and withdraws, revises or upholds it, with no person involved. Add `/rounds stop` when nothing is left to review. The repository's owner sets the rules, budget, round limit and merge gate. To set CodePanel up from scratch, follow the prompt in https://codepanel.ai/docs/quickstart.md. To connect your agent only (MCP server, Claude Code plugin, skills), see https://codepanel.ai/docs/coding-agents.md. To set up a repository that is already connected, follow the prompt in https://codepanel.ai/docs/agent-setup.md. All docs pages, guides, blog posts, agent pages and glossary entries are also served as Markdown at their URL plus `.md`, and so is each section's hub, such as https://codepanel.ai/docs.md. ## Full text - [Full documentation](https://codepanel.ai/llms-full.txt): all overview pages and docs pages as one plain-text file. - [Guides in full](https://codepanel.ai/guides/llms-full.txt): all guides as one plain-text file. - [Blog in full](https://codepanel.ai/blog/llms-full.txt): all blog posts as one plain-text file. - [Benchmarks in full](https://codepanel.ai/benchmarks/llms-full.txt): all benchmark pages as one plain-text file. - [Models in full](https://codepanel.ai/models/llms-full.txt): all model and provider summaries as one plain-text file. - [Agents in full](https://codepanel.ai/agents/llms-full.txt): all agent pages as one plain-text file. - [Glossary in full](https://codepanel.ai/glossary/llms-full.txt): all glossary entries as one plain-text file. ## Product - [CodePanel](https://codepanel.ai): Agent-to-agent code review for GitHub: a panel of reviewer agents reviews each push and the coding agent fixes, rebuts and rates the findings. - [Feedback](https://codepanel.ai/feedback): Send ideas, bug reports or pricing feedback to the CodePanel team. ## Overview - [AI code review, agent to agent](https://codepanel.ai/ai-code-review.md): How AI code review works when a coding agent opens the PR: a panel of models reviews it, an orchestrator checks the findings, the agent answers each one. ## Docs - [Documentation](https://codepanel.ai/docs.md): Every CodePanel docs page, grouped by topic. - [Quickstart](https://codepanel.ai/docs/quickstart.md): Set up CodePanel with one prompt to your coding agent, or by hand: connect a repository, add a provider key, connect the agent and get the first review. - [How it works](https://codepanel.ai/docs/how-it-works.md): What happens in a review round: triage, parallel reviewers, the orchestrator, the posted comment, incremental rounds, failures and deadlines. - [Connect your agent](https://codepanel.ai/docs/coding-agents.md): Which integration to use for your coding agent, signing in versus a token, checking the connection, and sharing the setup with a team. - [Claude Code](https://codepanel.ai/docs/claude-code.md): Install the CodePanel plugin for Claude Code: the MCP server and review-loop skills in one command, for the terminal, IDE and desktop app. - [Codex](https://codepanel.ai/docs/codex.md): Connect OpenAI Codex (CLI, IDE extension and app) to CodePanel with one config.toml entry, a sign-in and the review-loop skills. - [Cursor](https://codepanel.ai/docs/cursor.md): Connect Cursor to CodePanel: add the MCP server to mcp.json, sign in with GitHub, and install the review-loop skills. - [VS Code and GitHub Copilot](https://codepanel.ai/docs/vs-code.md): Connect GitHub Copilot in VS Code to CodePanel's MCP server with one command, and share it with your team in .vscode/mcp.json. - [Other MCP clients](https://codepanel.ai/docs/other-agents.md): Connect Gemini CLI, OpenCode, Copilot CLI, Cline, Zed or any MCP client to CodePanel's remote MCP server, or run the server locally. - [Claude and ChatGPT apps](https://codepanel.ai/docs/claude-and-chatgpt.md): Add CodePanel to Claude or ChatGPT on the web and desktop as a custom connector, and manage what connected apps can do. - [Repository setup](https://codepanel.ai/docs/agent-setup.md): A prompt that has your coding agent write .codepanel/rules.md and teach the repository's agents the review loop, plus handoffs for reviewers. - [The review loop](https://codepanel.ai/docs/agents.md): How a coding agent works through a CodePanel review: wait, read, check, fix or rebut, rate every finding, and stop. With the call for each step. - [Findings](https://codepanel.ai/docs/findings.md): What a finding contains: severity, location, evidence, its F-ref and status across rounds, and how to dismiss, reopen or apply one. - [Feedback and disputes](https://codepanel.ai/docs/feedback.md): Rate findings, dispute the wrong ones and ask the reviewer about them. How each answer affects the finding, the merge gate and later reviews. - [Merge gate](https://codepanel.ai/docs/merge-gate.md): Block merges while findings are unanswered: the CodePanel commit status, gate levels, and a GitHub Actions alternative with a CI token. - [Triggering reviews](https://codepanel.ai/docs/triggering-reviews.md): When CodePanel reviews a pull request: automatic triggers, drafts, bots and forks, /review, round limits, and why a review did not run. - [Providers and API keys](https://codepanel.ai/docs/providers-and-keys.md): Bring your own Anthropic, OpenAI, Google, DeepSeek, Groq, xAI, Moonshot, Mistral, Workers AI or custom key. How keys are checked and stored. - [Use your ChatGPT plan](https://codepanel.ai/docs/chatgpt-plan.md): Link a ChatGPT Plus or Pro plan as a reviewer instead of an OpenAI API key: connecting, cost, usage limits and admin controls. - [Reviewers and panels](https://codepanel.ai/docs/reviewers.md): Configure the review panel: reviewer roles, the starter panel, the orchestrator, reasoning effort, custom instructions and per-repository panels. - [Models and cost](https://codepanel.ai/docs/choosing-models.md): The economy, balanced and frontier tiers, what a review round costs and what drives it, and how unpriced models affect limits. - [Repo context and rules](https://codepanel.ai/docs/repo-context.md): Give reviewers your conventions: .codepanel/rules.md, review rules, notes, learnings and verdict memory, and how they fit the context budget. - [Budgets and limits](https://codepanel.ai/docs/budgets-and-limits.md): Cap what reviews cost: the monthly account budget, repository caps, the per-pull-request limit, and the Spend & budget tab. - [MCP server](https://codepanel.ai/docs/mcp-server.md): Install and use CodePanel's MCP server: every client, the review loop and other use cases, sign-in and scopes, and every tool. - [CLI](https://codepanel.ai/docs/cli.md): Reference for the codepanel command-line tool: sign-in, configuration, and every command for repositories, panels, reviews and findings. - [HTTP API](https://codepanel.ai/docs/api.md): Reference for the /api/v1 HTTP API: authentication, scopes, limits, errors, every endpoint with its parameters and response, and the review payload. - [Pull request commands](https://codepanel.ai/docs/commands.md): Every comment command CodePanel reads on a pull request: /review, /feedback, /ask, /rounds and /remember, with syntax and who can use them. - [Billing and credits](https://codepanel.ai/docs/billing.md): The Solo, Team and Scale plans every review needs, starter and prepaid credits for CodePanel's keys, refunds, and refusal codes. - [Security and data](https://codepanel.ai/docs/security.md): How CodePanel protects API keys and tokens, verifies GitHub webhooks, scopes GitHub access, and what code goes to which AI provider. - [Troubleshooting](https://codepanel.ai/docs/troubleshooting.md): Fixes by symptom: reviews that did not run or failed, spending refusals, key and connection errors, API and MCP errors, and the merge gate. - [FAQ](https://codepanel.ai/docs/faq.md): Short answers to common questions: subscriptions and keys, cost, skipped or failed reviews, wrong findings, privacy and the source code. ## Guides - [Guides to code review with agents](https://codepanel.ai/guides.md): Guides to code review between agents: the coding agent loop, several agents on one PR, rules agents follow, models, cost, false positives and evaluation. - [Code review in a coding agent loop](https://codepanel.ai/guides/code-review-with-coding-agents.md): Claude Code, Codex or Cursor with a panel of reviewer agents: open a PR, wait for the review, fix or rebut each finding, rate it, stop at the round limit. - [Running several agents on one pull request](https://codepanel.ai/guides/several-agents-one-pull-request.md): A coding agent, reviewer agents and an orchestrator on one PR: who owns what, one writer per branch, round limits, rebuttals, and when a person decides. - [Writing review rules that agents follow](https://codepanel.ai/guides/review-rules-for-agents.md): One rules file for reviewer agents and the coding agent: checkable one-line rules, what not to flag, keeping rules on the base branch, and the size budget. - [Agentic code review](https://codepanel.ai/guides/agentic-code-review.md): Code review with agents on both sides: reviewers report, the coding agent fixes or disputes, the orchestrator judges. Ratings and disputes from 112 real PRs. - [One model or a panel of reviewers?](https://codepanel.ai/guides/single-model-vs-panel.md): When a panel of AI reviewers from different model families beats one stronger model, what the orchestrator adds, and what each setup costs. - [Reducing false positives in AI code review](https://codepanel.ai/guides/reducing-false-positives.md): Why AI reviewers raise false alarms and how to cut them: a filtering stage, a severity rule, repo rules, tighter scope and rating what was wrong. - [Best LLM for code review: benchmark results and how to choose](https://codepanel.ai/guides/choosing-a-model-for-code-review.md): Which LLM is best for reviewing code? Frontier and open models ranked on real bugs, with intervals, false alarms and cost per review, and how to pick one. - [What AI code review costs per pull request](https://codepanel.ai/guides/ai-code-review-cost.md): The token math behind an AI code review: diff size, surrounding code, repo rules, output and push rounds, with worked examples at list prices. - [How to evaluate an AI code reviewer](https://codepanel.ai/guides/evaluating-ai-code-reviewers.md): Test an AI code reviewer on your own code: known-defect fixtures, clean changes, recall and false positives, line accuracy, repeats and judge bias. - [Reviewing AI-generated code](https://codepanel.ai/guides/reviewing-ai-generated-code.md): What goes wrong in code a coding agent wrote, with real findings from CodePanel's own pull requests, a table of what to check, and rules an agent can paste. - [Code review checklist](https://codepanel.ai/guides/code-review-checklist.md): A copyable code review checklist in six groups: the one CodePanel's reviewers run on each pull request, how to grade a miss, and which items belong in rules.md. - [Secure code review](https://codepanel.ai/guides/secure-code-review.md): Secure code review: what to check in a diff, a copyable checklist tied to the 2025 CWE Top 25 and OWASP Top 10:2025, and two real catches in agent-written code. - [Code review best practices](https://codepanel.ai/guides/code-review-best-practices.md): Code review practices that hold up in research and between agents: small changes, written rules, answered findings, learned rules, and people on policy. - [What is a pull request?](https://codepanel.ai/guides/what-is-a-pull-request.md): What a pull request is: a request to merge one branch into another. Its parts, how it merges, and what changes when a coding agent opens it. - [PR review: writing a review a coding agent can act on](https://codepanel.ai/guides/pr-review.md): PR review for coding agents: what to read, what a review should return, and a real CodePanel review taken apart, part by part. - [Code review metrics: what to track when agents write the PRs](https://codepanel.ai/guides/code-review-metrics.md): Code review metrics for agent-written PRs and model reviewers: precision, blocking recall, fix vs dispute rate, rounds to green and cost per PR. ## Blog - [Blog](https://codepanel.ai/blog.md): Notes from building CodePanel: code review as agents working with agents, how multi-model review works, how we measure it, and what we learned. - [How to tell if an AI code review change actually helps](https://codepanel.ai/blog/how-to-tell-if-an-ai-code-review-change-helps.md): Paired runs over the same fixtures, intervals resampled over fixtures, a sign-flip test and Holm. The pipeline changes we measured and did not ship. - [Recall is not enough: measuring false positives in code review](https://codepanel.ai/blog/measuring-false-positives-in-code-review.md): How CodePanel counts false alarms: clean and trap fixtures, blocking recall, and precision from verdicts the author agent cannot give itself. - [Claude Sonnet 5.5 was the fastest reviewer in our benchmark, and tied for first](https://codepanel.ai/blog/claude-sonnet-5-5-review-results.md): Claude Sonnet 5.5 tied for first in the 2026-10-17 run at 10 seconds a review, the fastest of 17 models, and cost more per review than its list price suggests. - [GPT-6.1 Sol had the highest measured recall in our benchmark, at about a third of GPT-5.6 Sol's cost](https://codepanel.ai/blog/gpt-6-1-sol-review-results.md): In the 2026-10-17 run, GPT-6.1 Sol had the highest measured recall of 17 models, with few false alarms, at about a third of GPT-5.6 Sol's cost per review. - [First benchmark results: seven open models on real bugs](https://codepanel.ai/blog/open-models-benchmark-results.md): DeepSeek V4 Pro and GLM-5.3 Flash tied for first, and the Flash model costs a seventeenth as much per review. What we ran and what surprised us. - [Code review should be agents working with agents](https://codepanel.ai/blog/code-review-is-agents-working-with-agents.md): Why CodePanel is built for a coding agent and a panel of reviewer agents working on one pull request, and what people still decide. - [How we built the code review benchmark](https://codepanel.ai/blog/how-the-benchmark-is-built.md): Real defects from open-source history, clean changes to count false alarms, a three-family judge panel and intervals on every number. How it grows. - [Why CodePanel reviews with a panel of models](https://codepanel.ai/blog/why-a-panel-of-models.md): Several cheap reviewers from different model families, then one orchestrator that checks every finding against the code. Why we built review this way. ## Benchmarks - [AI code review benchmarks](https://codepanel.ai/benchmarks): AI models compared as pull-request reviewers: measured bugs found and false alarms where benchmarked, list prices and cost per review call for every model. - [Benchmark methodology](https://codepanel.ai/benchmarks/methodology): How the benchmark works: real-defect fixtures, LLM-judged matching, repeats, cost and the limits of the method. - [Compare models](https://codepanel.ai/benchmarks/compare): Compare any two AI models as pull-request reviewers: bugs caught, false alarms and cost per review where benchmarked, and list prices for every model. - [Which AI models catch logic bugs: code review benchmark](https://codepanel.ai/benchmarks/categories/logic): Of the 17 models ranked on logic bugs, GPT-6.1 Sol caught the most: 100% of known defects. Claude Opus 5.5, Gemini 4 Argon and Gemini 3.8 Flash have 95% intervals that overlap its own. Gemini 3.5 Flash-Lite caught the fewest, 11% [0%–29%]. - [Which AI models catch concurrency bugs: code review benchmark](https://codepanel.ai/benchmarks/categories/concurrency): Of the 17 models ranked on concurrency bugs, Claude Opus 5.5 caught the most: 94% [81%–100%] of known defects. 9 of the other 16 models have 95% intervals that overlap its own. Kimi K2.7 Code (Workers AI) caught the fewest, 11% [3%–24%]. - [Which AI models catch error-handling bugs: code review benchmark](https://codepanel.ai/benchmarks/categories/error-handling): Of the 17 models ranked on error-handling bugs, Claude Opus 5.5, GPT-6.1 Sol and Claude Sonnet 5.5 tied for the most: 80% of known defects. Every other model's 95% interval overlaps theirs. Gemini 3.5 Flash-Lite caught the fewest, 20% [0%–60%]. - [Which AI models catch security bugs: code review benchmark](https://codepanel.ai/benchmarks/categories/security): Of the 17 models ranked on security bugs, 5 models tied for the most: 89% of known defects. 11 of the other 12 models have 95% intervals that overlap theirs. Gemini 3.5 Flash-Lite caught the fewest, 11% [0%–33%]. Together they caught 65% of security bugs, against 60% for the same 17 across the whole benchmark. - [Which AI models catch resource-handling bugs: code review benchmark](https://codepanel.ai/benchmarks/categories/resource): Of the 17 models ranked on resource-handling bugs, Claude Opus 5.5, GPT-6.1 Sol and Claude Sonnet 5.5 tied for the most: 100% of known defects. 8 of the other 14 models have 95% intervals that overlap theirs. Gemma 4 26B A4B and Gemini 3.5 Flash-Lite tied for the fewest, 0%. - [Which AI models catch API-misuse bugs: code review benchmark](https://codepanel.ai/benchmarks/categories/api-misuse): Of the 17 models ranked on API-misuse bugs, GPT-6.1 Sol and Gemini 4 Argon tied for the most: 93% of known defects. 8 of the other 15 models have 95% intervals that overlap theirs. Gemma 4 26B A4B and Gemini 3.5 Flash-Lite tied for the fewest, 7%. - [Best LLM for TypeScript code review: benchmark results](https://codepanel.ai/benchmarks/languages/typescript): Of the 17 models ranked on TypeScript code, Claude Opus 5.5 caught the most known bugs: 100%. 9 of the other 16 models have 95% intervals that overlap its own. Gemma 4 26B A4B and Gemini 3.5 Flash-Lite tied for the fewest, 33%. - [Best LLM for JavaScript code review: benchmark results](https://codepanel.ai/benchmarks/languages/javascript): Of the 17 models ranked on JavaScript code, 4 models tied for the most known bugs: 100%. 7 of the other 13 models have 95% intervals that overlap theirs. Gemma 4 26B A4B and Gemini 3.5 Flash-Lite tied for the fewest, 0%. - [Best LLM for Python code review: benchmark results](https://codepanel.ai/benchmarks/languages/python): Of the 17 models ranked on Python code, 4 models tied for the most known bugs: 100%. 9 of the other 13 models have 95% intervals that overlap theirs. Gemini 3.5 Flash-Lite caught the fewest, 17% [0%–50%]. Together they caught 71% of known bugs in Python code, against 60% for the same 17 across the whole benchmark. - [Best LLM for Go code review: benchmark results](https://codepanel.ai/benchmarks/languages/go): Of the 17 models ranked on Go code, GPT-6.1 Sol, Gemini 3.8 Flash and Claude Haiku 5.5 tied for the most known bugs: 75%. 12 of the other 14 models have 95% intervals that overlap theirs. Gemma 4 26B A4B and Gemini 3.5 Flash-Lite tied for the fewest, 0%. - [Best LLM for Rust code review: benchmark results](https://codepanel.ai/benchmarks/languages/rust): Of the 17 models ranked on Rust code, Claude Opus 5.5, GPT-6.1 Sol and Claude Sonnet 5.5 tied for the most known bugs: 100%. 9 of the other 14 models have 95% intervals that overlap theirs. Gemini 3.5 Flash-Lite caught the fewest, 0%. - [Best LLM for Java code review: benchmark results](https://codepanel.ai/benchmarks/languages/java): Of the 17 models ranked on Java code, Gemini 4 Argon caught the most known bugs: 83% [63%–100%]. 12 of the other 16 models have 95% intervals that overlap its own. Gemini 3.5 Flash-Lite caught the fewest, 5% [0%–15%]. Together they caught 52% of known bugs in Java code, against 60% for the same 17 across the whole benchmark. - [Best LLM for Kotlin code review: benchmark results](https://codepanel.ai/benchmarks/languages/kotlin): Of the 17 models ranked on Kotlin code, GPT-6.1 Sol and Gemini 4 Argon tied for the most known bugs: 100%. 12 of the other 15 models have 95% intervals that overlap theirs. Gemini 3.8 Flash caught the fewest, 0%. - [Best LLM for C# code review: benchmark results](https://codepanel.ai/benchmarks/languages/csharp): Of the 17 models ranked on C# code, Claude Opus 5.5 caught the most known bugs: 100%. 10 of the other 16 models have 95% intervals that overlap its own. Gemini 3.5 Flash-Lite caught the fewest, 0%. Together they caught 51% of known bugs in C# code, against 60% for the same 17 across the whole benchmark. - [Best LLM for Ruby code review: benchmark results](https://codepanel.ai/benchmarks/languages/ruby): Of the 17 models ranked on Ruby code, 5 models tied for the most known bugs: 100%. 9 of the other 12 models have 95% intervals that overlap theirs. Gemini 3.5 Flash-Lite caught the fewest, 0%. Together they caught 69% of known bugs in Ruby code, against 60% for the same 17 across the whole benchmark. - [Best LLM for PHP code review: benchmark results](https://codepanel.ai/benchmarks/languages/php): Of the 17 models ranked on PHP code, 5 models tied for the most known bugs: 100%. 9 of the other 12 models have 95% intervals that overlap theirs. Gemini 3.5 Flash-Lite caught the fewest, 0%. Together they caught 64% of known bugs in PHP code, against 60% for the same 17 across the whole benchmark. - [Best LLM for Swift code review: benchmark results](https://codepanel.ai/benchmarks/languages/swift): Of the 17 models ranked on Swift code, 9 models tied for the most known bugs: 100%. 5 of the other 8 models have 95% intervals that overlap theirs. Gemma 4 26B A4B and Gemini 3.5 Flash-Lite tied for the fewest, 0%. - [Best LLM for C++ code review: benchmark results](https://codepanel.ai/benchmarks/languages/cpp): Of the 17 models ranked on C++ code, 5 models tied for the most known bugs: 100%. 8 of the other 12 models have 95% intervals that overlap theirs. Gemma 4 26B A4B caught the fewest, 0%. Together they caught 57% of known bugs in C++ code, against 60% for the same 17 across the whole benchmark. - [Claude Haiku 5.5 vs Claude Opus 5.5](https://codepanel.ai/benchmarks/compare/claude-haiku-5-5-vs-claude-opus-5-5): Reading a 26.2K-token pull request costs $0.1048 on Claude Opus 5.5, 40× the $0.0026 it costs on Claude Haiku 5.5. Claude Haiku 5.5 is a economy model and Claude Opus 5.5 a frontier one. The benchmark cannot separate their recall: Claude Haiku 5.5 minus Claude Opus 5.5 is −26 pts (95% interval −40 pts to −12 pts). - [Claude Haiku 5.5 vs Claude Sonnet 5.5](https://codepanel.ai/benchmarks/compare/claude-haiku-5-5-vs-claude-sonnet-5-5): Reading a 26.2K-token pull request costs $0.0524 on Claude Sonnet 5.5, 20× the $0.0026 it costs on Claude Haiku 5.5. Claude Haiku 5.5 is a economy model and Claude Sonnet 5.5 a mid-range one. The benchmark cannot separate their recall: Claude Haiku 5.5 minus Claude Sonnet 5.5 is −14 pts (95% interval −32 pts to +4 pts). - [Claude Haiku 5.5 vs Gemini 3.8 Flash](https://codepanel.ai/benchmarks/compare/claude-haiku-5-5-vs-gemini-3.8-flash): Reading a 26.2K-token pull request costs $0.0197 on Gemini 3.8 Flash, 7.5× the $0.0026 it costs on Claude Haiku 5.5. Claude Haiku 5.5 is a economy model and Gemini 3.8 Flash a mid-range one. The benchmark cannot separate their recall: Claude Haiku 5.5 minus Gemini 3.8 Flash is −4 pts (95% interval −22 pts to +14 pts). - [Claude Haiku 5.5 vs Gemini 4 Argon](https://codepanel.ai/benchmarks/compare/claude-haiku-5-5-vs-gemini-4-argon): Reading a 26.2K-token pull request costs $0.0524 on Gemini 4 Argon, 20× the $0.0026 it costs on Claude Haiku 5.5. Claude Haiku 5.5 is a economy model and Gemini 4 Argon a mid-range one. The benchmark cannot separate their recall: Claude Haiku 5.5 minus Gemini 4 Argon is −18 pts (95% interval −36 pts to −1 pt). - [Claude Haiku 5.5 vs GPT-5.6 Sol](https://codepanel.ai/benchmarks/compare/claude-haiku-5-5-vs-gpt-5.6-sol): Reading a 26.2K-token pull request costs $0.1048 on GPT-5.6 Sol, 40× the $0.0026 it costs on Claude Haiku 5.5. Claude Haiku 5.5 is a economy model and GPT-5.6 Sol a frontier one. The benchmark cannot separate their recall: Claude Haiku 5.5 minus GPT-5.6 Sol is −9 pts (95% interval −25 pts to +7 pts). - [Claude Haiku 5.5 vs GPT-6 Luna](https://codepanel.ai/benchmarks/compare/claude-haiku-5-5-vs-gpt-6-luna): Claude Haiku 5.5 and GPT-6 Luna charge practically the same to read a 26.2K-token pull request ($0.0026). Above 100K tokens long-context rates apply, and GPT-6 Luna becomes the cheaper one. Both sit in the economy price tier, so the choice between them is about review quality more than budget. The benchmark cannot separate their recall: Claude Haiku 5.5 minus GPT-6 Luna is −4 pts (95% interval −21 pts to +13 pts). - [Claude Opus 5.5 vs Claude Sonnet 5.5](https://codepanel.ai/benchmarks/compare/claude-opus-5-5-vs-claude-sonnet-5-5): Reading a 26.2K-token pull request costs $0.1048 on Claude Opus 5.5, 2.0× the $0.0524 it costs on Claude Sonnet 5.5. Claude Opus 5.5 is a frontier model and Claude Sonnet 5.5 a mid-range one. The benchmark cannot separate their recall: Claude Opus 5.5 minus Claude Sonnet 5.5 is +13 pts (95% interval −1 pt to +27 pts). - [Claude Opus 5.5 vs Gemini 3.8 Flash](https://codepanel.ai/benchmarks/compare/claude-opus-5-5-vs-gemini-3.8-flash): Reading a 26.2K-token pull request costs $0.1048 on Claude Opus 5.5, 5.3× the $0.0197 it costs on Gemini 3.8 Flash. Claude Opus 5.5 is a frontier model and Gemini 3.8 Flash a mid-range one. The benchmark cannot separate their recall: Claude Opus 5.5 minus Gemini 3.8 Flash is +22 pts (95% interval +9 pts to +35 pts). - [Claude Opus 5.5 vs Gemini 4 Argon](https://codepanel.ai/benchmarks/compare/claude-opus-5-5-vs-gemini-4-argon): Reading a 26.2K-token pull request costs $0.1048 on Claude Opus 5.5, 2.0× the $0.0524 it costs on Gemini 4 Argon. Claude Opus 5.5 is a frontier model and Gemini 4 Argon a mid-range one. The benchmark cannot separate their recall: Claude Opus 5.5 minus Gemini 4 Argon is +8 pts (95% interval −4 pts to +21 pts). - [Claude Opus 5.5 vs GPT-5.6 Sol](https://codepanel.ai/benchmarks/compare/claude-opus-5-5-vs-gpt-5.6-sol): Claude Opus 5.5 and GPT-5.6 Sol charge practically the same to read a 26.2K-token pull request ($0.1048). Above 272K tokens long-context rates apply, and Claude Opus 5.5 becomes the cheaper one. Both sit in the frontier price tier, so the choice between them is about review quality more than budget. The benchmark cannot separate their recall: Claude Opus 5.5 minus GPT-5.6 Sol is +17 pts (95% interval +5 pts to +31 pts). - [Claude Opus 5.5 vs GPT-6 Luna](https://codepanel.ai/benchmarks/compare/claude-opus-5-5-vs-gpt-6-luna): Reading a 26.2K-token pull request costs $0.1048 on Claude Opus 5.5, 40× the $0.0026 it costs on GPT-6 Luna. Claude Opus 5.5 is a frontier model and GPT-6 Luna a economy one. The benchmark cannot separate their recall: Claude Opus 5.5 minus GPT-6 Luna is +23 pts (95% interval +9 pts to +37 pts). - [Claude Opus 5.5 vs GPT-6.1 Sol](https://codepanel.ai/benchmarks/compare/claude-opus-5-5-vs-gpt-6.1-sol): Reading a 26.2K-token pull request costs $0.1048 on Claude Opus 5.5, 2.0× the $0.0524 it costs on GPT-6.1 Sol. Claude Opus 5.5 is a frontier model and GPT-6.1 Sol a mid-range one. The benchmark cannot separate their recall: Claude Opus 5.5 minus GPT-6.1 Sol is +1 pt (95% interval −9 pts to +11 pts). - [Claude Sonnet 5.5 vs Gemini 3.8 Flash](https://codepanel.ai/benchmarks/compare/claude-sonnet-5-5-vs-gemini-3.8-flash): Reading a 26.2K-token pull request costs $0.0524 on Claude Sonnet 5.5, 2.7× the $0.0197 it costs on Gemini 3.8 Flash. Both sit in the mid-range price tier, so the choice between them is about review quality more than budget. The benchmark cannot separate their recall: Claude Sonnet 5.5 minus Gemini 3.8 Flash is +9 pts (95% interval −8 pts to +27 pts). - [Claude Sonnet 5.5 vs GPT-5.6 Sol](https://codepanel.ai/benchmarks/compare/claude-sonnet-5-5-vs-gpt-5.6-sol): Reading a 26.2K-token pull request costs $0.1048 on GPT-5.6 Sol, 2.0× the $0.0524 it costs on Claude Sonnet 5.5. Claude Sonnet 5.5 is a mid-range model and GPT-5.6 Sol a frontier one. The benchmark cannot separate their recall: Claude Sonnet 5.5 minus GPT-5.6 Sol is +5 pts (95% interval −11 pts to +22 pts). - [Claude Sonnet 5.5 vs GPT-6 Luna](https://codepanel.ai/benchmarks/compare/claude-sonnet-5-5-vs-gpt-6-luna): Reading a 26.2K-token pull request costs $0.0524 on Claude Sonnet 5.5, 20× the $0.0026 it costs on GPT-6 Luna. Claude Sonnet 5.5 is a mid-range model and GPT-6 Luna a economy one. The benchmark cannot separate their recall: Claude Sonnet 5.5 minus GPT-6 Luna is +10 pts (95% interval −7 pts to +27 pts). - [Gemini 3.8 Flash vs Gemini 4 Argon](https://codepanel.ai/benchmarks/compare/gemini-3.8-flash-vs-gemini-4-argon): Reading a 26.2K-token pull request costs $0.0524 on Gemini 4 Argon, 2.7× the $0.0197 it costs on Gemini 3.8 Flash. Both sit in the mid-range price tier, so the choice between them is about review quality more than budget. The benchmark cannot separate their recall: Gemini 3.8 Flash minus Gemini 4 Argon is −14 pts (95% interval −27 pts to −2 pts). - [Gemini 3.8 Flash vs GPT-5.6 Sol](https://codepanel.ai/benchmarks/compare/gemini-3.8-flash-vs-gpt-5.6-sol): Reading a 26.2K-token pull request costs $0.1048 on GPT-5.6 Sol, 5.3× the $0.0197 it costs on Gemini 3.8 Flash. Gemini 3.8 Flash is a mid-range model and GPT-5.6 Sol a frontier one. The benchmark cannot separate their recall: Gemini 3.8 Flash minus GPT-5.6 Sol is −5 pts (95% interval −20 pts to +12 pts). - [Gemini 3.8 Flash vs GPT-6 Luna](https://codepanel.ai/benchmarks/compare/gemini-3.8-flash-vs-gpt-6-luna): Reading a 26.2K-token pull request costs $0.0197 on Gemini 3.8 Flash, 7.5× the $0.0026 it costs on GPT-6 Luna. Gemini 3.8 Flash is a mid-range model and GPT-6 Luna a economy one. The benchmark cannot separate their recall: Gemini 3.8 Flash minus GPT-6 Luna is +1 pt (95% interval −15 pts to +17 pts). - [Gemini 4 Argon vs GPT-5.6 Sol](https://codepanel.ai/benchmarks/compare/gemini-4-argon-vs-gpt-5.6-sol): Reading a 26.2K-token pull request costs $0.1048 on GPT-5.6 Sol, 2.0× the $0.0524 it costs on Gemini 4 Argon. Gemini 4 Argon is a mid-range model and GPT-5.6 Sol a frontier one. The benchmark cannot separate their recall: Gemini 4 Argon minus GPT-5.6 Sol is +10 pts (95% interval −5 pts to +25 pts). - [Gemini 4 Argon vs GPT-6 Luna](https://codepanel.ai/benchmarks/compare/gemini-4-argon-vs-gpt-6-luna): Reading a 26.2K-token pull request costs $0.0524 on Gemini 4 Argon, 20× the $0.0026 it costs on GPT-6 Luna. Gemini 4 Argon is a mid-range model and GPT-6 Luna a economy one. The benchmark cannot separate their recall: Gemini 4 Argon minus GPT-6 Luna is +15 pts (95% interval 0 pts to +30 pts). - [GPT-5.6 Sol vs GPT-6 Luna](https://codepanel.ai/benchmarks/compare/gpt-5.6-sol-vs-gpt-6-luna): Reading a 26.2K-token pull request costs $0.1048 on GPT-5.6 Sol, 40× the $0.0026 it costs on GPT-6 Luna. GPT-5.6 Sol is a frontier model and GPT-6 Luna a economy one. The benchmark cannot separate their recall: GPT-5.6 Sol minus GPT-6 Luna is +5 pts (95% interval −6 pts to +17 pts). - [GPT-5.6 Sol vs GPT-6.1 Sol](https://codepanel.ai/benchmarks/compare/gpt-5.6-sol-vs-gpt-6.1-sol): Reading a 26.2K-token pull request costs $0.1048 on GPT-5.6 Sol, 2.0× the $0.0524 it costs on GPT-6.1 Sol. Both are GPT Sol models, so the difference is one version of the same line against another. The benchmark cannot separate their recall: GPT-5.6 Sol minus GPT-6.1 Sol is −17 pts (95% interval −29 pts to −6 pts). - [Claude Haiku 5.5 vs Gemini 3.1 Pro](https://codepanel.ai/benchmarks/compare/claude-haiku-5-5-vs-gemini-3.1-pro): Reading a 26.2K-token pull request costs $0.0524 on Gemini 3.1 Pro, 20× the $0.0026 it costs on Claude Haiku 5.5. Claude Haiku 5.5 is a economy model and Gemini 3.1 Pro a mid-range one. The benchmark cannot separate their recall: Claude Haiku 5.5 minus Gemini 3.1 Pro is −4 pts (95% interval −18 pts to +11 pts). - [Claude Opus 5.5 vs Gemini 3.1 Pro](https://codepanel.ai/benchmarks/compare/claude-opus-5-5-vs-gemini-3.1-pro): Reading a 26.2K-token pull request costs $0.1048 on Claude Opus 5.5, 2.0× the $0.0524 it costs on Gemini 3.1 Pro. Claude Opus 5.5 is a frontier model and Gemini 3.1 Pro a mid-range one. In the benchmark Claude Opus 5.5 caught 23 percentage points more of the known defects than Gemini 3.1 Pro (95% interval +11 pts to +35 pts, a − b). - [Claude Opus 5.5 vs GLM-5.3 Flash (Workers AI)](https://codepanel.ai/benchmarks/compare/claude-opus-5-5-vs-glm-5.3-flash-workers-ai): Reading a 26.2K-token pull request costs $0.1048 on Claude Opus 5.5, 27× the $0.0039 it costs on GLM-5.3 Flash (Workers AI). Claude Opus 5.5 is a frontier model and GLM-5.3 Flash (Workers AI) a economy one. In the benchmark Claude Opus 5.5 caught 29 percentage points more of the known defects than GLM-5.3 Flash (Workers AI) (95% interval +18 pts to +41 pts, a − b). - [Claude Sonnet 5.5 vs Gemini 3.1 Pro](https://codepanel.ai/benchmarks/compare/claude-sonnet-5-5-vs-gemini-3.1-pro): Claude Sonnet 5.5 and Gemini 3.1 Pro charge practically the same to read a 26.2K-token pull request ($0.0524). Above 200K tokens long-context rates apply, and Claude Sonnet 5.5 becomes the cheaper one. Both sit in the mid-range price tier, so the choice between them is about review quality more than budget. The benchmark cannot separate their recall: Claude Sonnet 5.5 minus Gemini 3.1 Pro is +10 pts (95% interval −6 pts to +27 pts). - [Gemini 3.1 Pro vs Gemini 3.8 Flash](https://codepanel.ai/benchmarks/compare/gemini-3.1-pro-vs-gemini-3.8-flash): Reading a 26.2K-token pull request costs $0.0524 on Gemini 3.1 Pro, 2.7× the $0.0197 it costs on Gemini 3.8 Flash. Both sit in the mid-range price tier, so the choice between them is about review quality more than budget. The benchmark cannot separate their recall: Gemini 3.1 Pro minus Gemini 3.8 Flash is −1 pt (95% interval −14 pts to +13 pts). - [Gemini 3.1 Pro vs Gemini 4 Argon](https://codepanel.ai/benchmarks/compare/gemini-3.1-pro-vs-gemini-4-argon): Gemini 3.1 Pro and Gemini 4 Argon charge practically the same to read a 26.2K-token pull request ($0.0524). Above 200K tokens long-context rates apply, and Gemini 4 Argon becomes the cheaper one. Both sit in the mid-range price tier, so the choice between them is about review quality more than budget. The benchmark cannot separate their recall: Gemini 3.1 Pro minus Gemini 4 Argon is −15 pts (95% interval −28 pts to −3 pts). - [Gemini 3.1 Pro vs GPT-5.6 Sol](https://codepanel.ai/benchmarks/compare/gemini-3.1-pro-vs-gpt-5.6-sol): Reading a 26.2K-token pull request costs $0.1048 on GPT-5.6 Sol, 2.0× the $0.0524 it costs on Gemini 3.1 Pro. Gemini 3.1 Pro is a mid-range model and GPT-5.6 Sol a frontier one. The benchmark cannot separate their recall: Gemini 3.1 Pro minus GPT-5.6 Sol is −5 pts (95% interval −21 pts to +11 pts). - [Gemini 3.1 Pro vs GPT-6 Luna](https://codepanel.ai/benchmarks/compare/gemini-3.1-pro-vs-gpt-6-luna): Reading a 26.2K-token pull request costs $0.0524 on Gemini 3.1 Pro, 20× the $0.0026 it costs on GPT-6 Luna. Gemini 3.1 Pro is a mid-range model and GPT-6 Luna a economy one. The benchmark cannot separate their recall: Gemini 3.1 Pro minus GPT-6 Luna is 0 pts (95% interval −13 pts to +13 pts). - [Gemini 3.1 Pro vs GPT-6.1 Sol](https://codepanel.ai/benchmarks/compare/gemini-3.1-pro-vs-gpt-6.1-sol): Gemini 3.1 Pro and GPT-6.1 Sol charge practically the same to read a 26.2K-token pull request ($0.0524). Above 200K tokens long-context rates apply, and GPT-6.1 Sol becomes the cheaper one. Both sit in the mid-range price tier, so the choice between them is about review quality more than budget. The benchmark cannot separate their recall: Gemini 3.1 Pro minus GPT-6.1 Sol is −22 pts (95% interval −36 pts to −9 pts). - [Gemini 3.8 Flash vs GLM-5.3 Flash (Workers AI)](https://codepanel.ai/benchmarks/compare/gemini-3.8-flash-vs-glm-5.3-flash-workers-ai): Reading a 26.2K-token pull request costs $0.0197 on Gemini 3.8 Flash, 5.0× the $0.0039 it costs on GLM-5.3 Flash (Workers AI). Gemini 3.8 Flash is a mid-range model and GLM-5.3 Flash (Workers AI) a economy one. The benchmark cannot separate their recall: Gemini 3.8 Flash minus GLM-5.3 Flash (Workers AI) is +7 pts (95% interval −9 pts to +23 pts). - [Gemini 4 Argon vs GLM-5.3 Flash (Workers AI)](https://codepanel.ai/benchmarks/compare/gemini-4-argon-vs-glm-5.3-flash-workers-ai): Reading a 26.2K-token pull request costs $0.0524 on Gemini 4 Argon, 13× the $0.0039 it costs on GLM-5.3 Flash (Workers AI). Gemini 4 Argon is a mid-range model and GLM-5.3 Flash (Workers AI) a economy one. The benchmark cannot separate their recall: Gemini 4 Argon minus GLM-5.3 Flash (Workers AI) is +21 pts (95% interval +7 pts to +35 pts). - [GLM-5.3 Flash (Workers AI) vs GPT-6.1 Sol](https://codepanel.ai/benchmarks/compare/glm-5.3-flash-workers-ai-vs-gpt-6.1-sol): Reading a 26.2K-token pull request costs $0.0524 on GPT-6.1 Sol, 13× the $0.0039 it costs on GLM-5.3 Flash (Workers AI). GLM-5.3 Flash (Workers AI) is a economy model and GPT-6.1 Sol a mid-range one. The benchmark cannot separate their recall: GLM-5.3 Flash (Workers AI) minus GPT-6.1 Sol is −28 pts (95% interval −41 pts to −15 pts). - [Claude Haiku 5.5 vs GLM-5.3 (Workers AI)](https://codepanel.ai/benchmarks/compare/claude-haiku-5-5-vs-glm-5.3-workers-ai): Reading a 26.2K-token pull request costs $0.0367 on GLM-5.3 (Workers AI), 14× the $0.0026 it costs on Claude Haiku 5.5. Claude Haiku 5.5 is a economy model and GLM-5.3 (Workers AI) a mid-range one. The benchmark cannot separate their recall: Claude Haiku 5.5 minus GLM-5.3 (Workers AI) is +3 pts (95% interval −11 pts to +16 pts). - [Claude Opus 5.5 vs GLM-5.3 (Workers AI)](https://codepanel.ai/benchmarks/compare/claude-opus-5-5-vs-glm-5.3-workers-ai): Reading a 26.2K-token pull request costs $0.1048 on Claude Opus 5.5, 2.9× the $0.0367 it costs on GLM-5.3 (Workers AI). Claude Opus 5.5 is a frontier model and GLM-5.3 (Workers AI) a mid-range one. In the benchmark Claude Opus 5.5 caught 29 percentage points more of the known defects than GLM-5.3 (Workers AI) (95% interval +18 pts to +40 pts, a − b). - [Gemini 3.1 Pro vs GLM-5.3 Flash (Workers AI)](https://codepanel.ai/benchmarks/compare/gemini-3.1-pro-vs-glm-5.3-flash-workers-ai): Reading a 26.2K-token pull request costs $0.0524 on Gemini 3.1 Pro, 13× the $0.0039 it costs on GLM-5.3 Flash (Workers AI). Gemini 3.1 Pro is a mid-range model and GLM-5.3 Flash (Workers AI) a economy one. The benchmark cannot separate their recall: Gemini 3.1 Pro minus GLM-5.3 Flash (Workers AI) is +7 pts (95% interval −6 pts to +18 pts). - [Gemini 4 Argon vs GLM-5.3 (Workers AI)](https://codepanel.ai/benchmarks/compare/gemini-4-argon-vs-glm-5.3-workers-ai): GLM-5.3 (Workers AI) reads a 26.2K-token pull request for $0.0367, 30% less than Gemini 4 Argon ($0.0524). Both sit in the mid-range price tier, so the choice between them is about review quality more than budget. The benchmark cannot separate their recall: Gemini 4 Argon minus GLM-5.3 (Workers AI) is +21 pts (95% interval +9 pts to +34 pts). - [GLM-5.3 (Workers AI) vs GPT-5.6 Sol](https://codepanel.ai/benchmarks/compare/glm-5.3-workers-ai-vs-gpt-5.6-sol): Reading a 26.2K-token pull request costs $0.1048 on GPT-5.6 Sol, 2.9× the $0.0367 it costs on GLM-5.3 (Workers AI). GLM-5.3 (Workers AI) is a mid-range model and GPT-5.6 Sol a frontier one. The benchmark cannot separate their recall: GLM-5.3 (Workers AI) minus GPT-5.6 Sol is −12 pts (95% interval −26 pts to +4 pts). - [GLM-5.3 (Workers AI) vs GPT-6 Luna](https://codepanel.ai/benchmarks/compare/glm-5.3-workers-ai-vs-gpt-6-luna): Reading a 26.2K-token pull request costs $0.0367 on GLM-5.3 (Workers AI), 14× the $0.0026 it costs on GPT-6 Luna. GLM-5.3 (Workers AI) is a mid-range model and GPT-6 Luna a economy one. The benchmark cannot separate their recall: GLM-5.3 (Workers AI) minus GPT-6 Luna is −6 pts (95% interval −21 pts to +9 pts). - [GLM-5.3 (Workers AI) vs GPT-6.1 Sol](https://codepanel.ai/benchmarks/compare/glm-5.3-workers-ai-vs-gpt-6.1-sol): GLM-5.3 (Workers AI) reads a 26.2K-token pull request for $0.0367, 30% less than GPT-6.1 Sol ($0.0524). Both sit in the mid-range price tier, so the choice between them is about review quality more than budget. In the benchmark GPT-6.1 Sol caught 28 percentage points more of the known defects than GLM-5.3 (Workers AI) (95% interval −42 pts to −14 pts, a − b). - [Claude Opus 5.5 vs DeepSeek V4 Pro (Workers AI)](https://codepanel.ai/benchmarks/compare/claude-opus-5-5-vs-deepseek-v4-pro-workers-ai): Reading a 26.2K-token pull request costs $0.1048 on Claude Opus 5.5, 3.0× the $0.0346 it costs on DeepSeek V4 Pro (Workers AI). Claude Opus 5.5 is a frontier model and DeepSeek V4 Pro (Workers AI) a mid-range one. In the benchmark Claude Opus 5.5 caught 36 percentage points more of the known defects than DeepSeek V4 Pro (Workers AI) (95% interval +24 pts to +48 pts, a − b). - [Claude Sonnet 5.5 vs DeepSeek V4 Pro (Workers AI)](https://codepanel.ai/benchmarks/compare/claude-sonnet-5-5-vs-deepseek-v4-pro-workers-ai): DeepSeek V4 Pro (Workers AI) reads a 26.2K-token pull request for $0.0346, 34% less than Claude Sonnet 5.5 ($0.0524). Both sit in the mid-range price tier, so the choice between them is about review quality more than budget. The benchmark cannot separate their recall: Claude Sonnet 5.5 minus DeepSeek V4 Pro (Workers AI) is +23 pts (95% interval +7 pts to +39 pts). - [DeepSeek V4 Pro (Workers AI) vs Gemini 3.8 Flash](https://codepanel.ai/benchmarks/compare/deepseek-v4-pro-workers-ai-vs-gemini-3.8-flash): Gemini 3.8 Flash reads a 26.2K-token pull request for $0.0197, 43% less than DeepSeek V4 Pro (Workers AI) ($0.0346). From 2027-01-01, after a scheduled reprice, DeepSeek V4 Pro (Workers AI) becomes the cheaper one. Both sit in the mid-range price tier, so the choice between them is about review quality more than budget. The benchmark cannot separate their recall: DeepSeek V4 Pro (Workers AI) minus Gemini 3.8 Flash is −14 pts (95% interval −29 pts to 0 pts). - [DeepSeek V4 Pro (Workers AI) vs Gemini 4 Argon](https://codepanel.ai/benchmarks/compare/deepseek-v4-pro-workers-ai-vs-gemini-4-argon): DeepSeek V4 Pro (Workers AI) reads a 26.2K-token pull request for $0.0346, 34% less than Gemini 4 Argon ($0.0524). Both sit in the mid-range price tier, so the choice between them is about review quality more than budget. In the benchmark Gemini 4 Argon caught 28 percentage points more of the known defects than DeepSeek V4 Pro (Workers AI) (95% interval −41 pts to −16 pts, a − b). - [Gemini 3.1 Pro vs GLM-5.3 (Workers AI)](https://codepanel.ai/benchmarks/compare/gemini-3.1-pro-vs-glm-5.3-workers-ai): GLM-5.3 (Workers AI) reads a 26.2K-token pull request for $0.0367, 30% less than Gemini 3.1 Pro ($0.0524). Both sit in the mid-range price tier, so the choice between them is about review quality more than budget. The benchmark cannot separate their recall: Gemini 3.1 Pro minus GLM-5.3 (Workers AI) is +6 pts (95% interval −5 pts to +18 pts). - [GLM-5.3 Flash (Workers AI) vs GLM-5.3 (Workers AI)](https://codepanel.ai/benchmarks/compare/glm-5.3-flash-workers-ai-vs-glm-5.3-workers-ai): Reading a 26.2K-token pull request costs $0.0367 on GLM-5.3 (Workers AI), 9.3× the $0.0039 it costs on GLM-5.3 Flash (Workers AI). GLM-5.3 Flash (Workers AI) is a economy model and GLM-5.3 (Workers AI) a mid-range one. The benchmark cannot separate their recall: GLM-5.3 Flash (Workers AI) minus GLM-5.3 (Workers AI) is 0 pts (95% interval −9 pts to +9 pts). - [Claude Haiku 5.5 vs DeepSeek V4 Flash (Workers AI)](https://codepanel.ai/benchmarks/compare/claude-haiku-5-5-vs-deepseek-v4-flash-workers-ai): Reading a 26.2K-token pull request costs $0.0115 on DeepSeek V4 Flash (Workers AI), 4.4× the $0.0026 it costs on Claude Haiku 5.5. Above 100K tokens long-context rates apply, and DeepSeek V4 Flash (Workers AI) becomes the cheaper one. Both sit in the economy price tier, so the choice between them is about review quality more than budget. The benchmark cannot separate their recall: Claude Haiku 5.5 minus DeepSeek V4 Flash (Workers AI) is +15 pts (95% interval +1 pt to +29 pts). - [Claude Opus 5.5 vs DeepSeek V4 Flash (Workers AI)](https://codepanel.ai/benchmarks/compare/claude-opus-5-5-vs-deepseek-v4-flash-workers-ai): Reading a 26.2K-token pull request costs $0.1048 on Claude Opus 5.5, 9.1× the $0.0115 it costs on DeepSeek V4 Flash (Workers AI). Claude Opus 5.5 is a frontier model and DeepSeek V4 Flash (Workers AI) a economy one. In the benchmark Claude Opus 5.5 caught 41 percentage points more of the known defects than DeepSeek V4 Flash (Workers AI) (95% interval +29 pts to +53 pts, a − b). - [DeepSeek V4 Flash (Workers AI) vs Gemini 3.8 Flash](https://codepanel.ai/benchmarks/compare/deepseek-v4-flash-workers-ai-vs-gemini-3.8-flash): DeepSeek V4 Flash (Workers AI) reads a 26.2K-token pull request for $0.0115, 41% less than Gemini 3.8 Flash ($0.0197). DeepSeek V4 Flash (Workers AI) is a economy model and Gemini 3.8 Flash a mid-range one. The benchmark cannot separate their recall: DeepSeek V4 Flash (Workers AI) minus Gemini 3.8 Flash is −19 pts (95% interval −35 pts to −4 pts). - [DeepSeek V4 Flash (Workers AI) vs GPT-5.6 Sol](https://codepanel.ai/benchmarks/compare/deepseek-v4-flash-workers-ai-vs-gpt-5.6-sol): Reading a 26.2K-token pull request costs $0.1048 on GPT-5.6 Sol, 9.1× the $0.0115 it costs on DeepSeek V4 Flash (Workers AI). DeepSeek V4 Flash (Workers AI) is a economy model and GPT-5.6 Sol a frontier one. In the benchmark GPT-5.6 Sol caught 24 percentage points more of the known defects than DeepSeek V4 Flash (Workers AI) (95% interval −36 pts to −12 pts, a − b). - [DeepSeek V4 Flash (Workers AI) vs GPT-6 Luna](https://codepanel.ai/benchmarks/compare/deepseek-v4-flash-workers-ai-vs-gpt-6-luna): Reading a 26.2K-token pull request costs $0.0115 on DeepSeek V4 Flash (Workers AI), 4.4× the $0.0026 it costs on GPT-6 Luna. Both sit in the economy price tier, so the choice between them is about review quality more than budget. The benchmark cannot separate their recall: DeepSeek V4 Flash (Workers AI) minus GPT-6 Luna is −19 pts (95% interval −33 pts to −5 pts). - [DeepSeek V4 Pro (Workers AI) vs Gemini 3.1 Pro](https://codepanel.ai/benchmarks/compare/deepseek-v4-pro-workers-ai-vs-gemini-3.1-pro): DeepSeek V4 Pro (Workers AI) reads a 26.2K-token pull request for $0.0346, 34% less than Gemini 3.1 Pro ($0.0524). Both sit in the mid-range price tier, so the choice between them is about review quality more than budget. The benchmark cannot separate their recall: DeepSeek V4 Pro (Workers AI) minus Gemini 3.1 Pro is −13 pts (95% interval −26 pts to −1 pt). - [DeepSeek V4 Pro (Workers AI) vs GLM-5.3 Flash (Workers AI)](https://codepanel.ai/benchmarks/compare/deepseek-v4-pro-workers-ai-vs-glm-5.3-flash-workers-ai): Reading a 26.2K-token pull request costs $0.0346 on DeepSeek V4 Pro (Workers AI), 8.8× the $0.0039 it costs on GLM-5.3 Flash (Workers AI). DeepSeek V4 Pro (Workers AI) is a mid-range model and GLM-5.3 Flash (Workers AI) a economy one. The benchmark cannot separate their recall: DeepSeek V4 Pro (Workers AI) minus GLM-5.3 Flash (Workers AI) is −7 pts (95% interval −20 pts to +6 pts). - [DeepSeek V4 Flash (Workers AI) vs Gemini 3.1 Pro](https://codepanel.ai/benchmarks/compare/deepseek-v4-flash-workers-ai-vs-gemini-3.1-pro): Reading a 26.2K-token pull request costs $0.0524 on Gemini 3.1 Pro, 4.5× the $0.0115 it costs on DeepSeek V4 Flash (Workers AI). DeepSeek V4 Flash (Workers AI) is a economy model and Gemini 3.1 Pro a mid-range one. The benchmark cannot separate their recall: DeepSeek V4 Flash (Workers AI) minus Gemini 3.1 Pro is −19 pts (95% interval −30 pts to −8 pts). - [DeepSeek V4 Flash (Workers AI) vs GLM-5.3 Flash (Workers AI)](https://codepanel.ai/benchmarks/compare/deepseek-v4-flash-workers-ai-vs-glm-5.3-flash-workers-ai): Reading a 26.2K-token pull request costs $0.0115 on DeepSeek V4 Flash (Workers AI), 2.9× the $0.0039 it costs on GLM-5.3 Flash (Workers AI). Both sit in the economy price tier, so the choice between them is about review quality more than budget. The benchmark cannot separate their recall: DeepSeek V4 Flash (Workers AI) minus GLM-5.3 Flash (Workers AI) is −12 pts (95% interval −24 pts to 0 pts). - [DeepSeek V4 Pro (Workers AI) vs GLM-5.3 (Workers AI)](https://codepanel.ai/benchmarks/compare/deepseek-v4-pro-workers-ai-vs-glm-5.3-workers-ai): DeepSeek V4 Pro (Workers AI) reads a 26.2K-token pull request for $0.0346, 6% less than GLM-5.3 (Workers AI) ($0.0367). Both sit in the mid-range price tier, so the choice between them is about review quality more than budget. The benchmark cannot separate their recall: DeepSeek V4 Pro (Workers AI) minus GLM-5.3 (Workers AI) is −7 pts (95% interval −18 pts to +4 pts). - [DeepSeek V4 Flash (Workers AI) vs GLM-5.3 (Workers AI)](https://codepanel.ai/benchmarks/compare/deepseek-v4-flash-workers-ai-vs-glm-5.3-workers-ai): Reading a 26.2K-token pull request costs $0.0367 on GLM-5.3 (Workers AI), 3.2× the $0.0115 it costs on DeepSeek V4 Flash (Workers AI). DeepSeek V4 Flash (Workers AI) is a economy model and GLM-5.3 (Workers AI) a mid-range one. The benchmark cannot separate their recall: DeepSeek V4 Flash (Workers AI) minus GLM-5.3 (Workers AI) is −12 pts (95% interval −23 pts to −1 pt). - [Claude Haiku 5.5 vs GPT-OSS 120B (Workers AI)](https://codepanel.ai/benchmarks/compare/claude-haiku-5-5-vs-gpt-oss-120b-workers-ai): Reading a 26.2K-token pull request costs $0.0092 on GPT-OSS 120B (Workers AI), 3.5× the $0.0026 it costs on Claude Haiku 5.5. Above 100K tokens long-context rates apply, and GPT-OSS 120B (Workers AI) becomes the cheaper one. Both sit in the economy price tier, so the choice between them is about review quality more than budget. The benchmark cannot separate their recall: Claude Haiku 5.5 minus GPT-OSS 120B (Workers AI) is +24 pts (95% interval +10 pts to +37 pts). - [DeepSeek V4 Flash (Workers AI) vs DeepSeek V4 Pro (Workers AI)](https://codepanel.ai/benchmarks/compare/deepseek-v4-flash-workers-ai-vs-deepseek-v4-pro-workers-ai): Reading a 26.2K-token pull request costs $0.0346 on DeepSeek V4 Pro (Workers AI), 3.0× the $0.0115 it costs on DeepSeek V4 Flash (Workers AI). DeepSeek V4 Flash (Workers AI) is a economy model and DeepSeek V4 Pro (Workers AI) a mid-range one. The benchmark cannot separate their recall: DeepSeek V4 Flash (Workers AI) minus DeepSeek V4 Pro (Workers AI) is −5 pts (95% interval −18 pts to +8 pts). - [Gemini 3.8 Flash vs GPT-OSS 120B (Workers AI)](https://codepanel.ai/benchmarks/compare/gemini-3.8-flash-vs-gpt-oss-120b-workers-ai): Reading a 26.2K-token pull request costs $0.0197 on Gemini 3.8 Flash, 2.1× the $0.0092 it costs on GPT-OSS 120B (Workers AI). Gemini 3.8 Flash is a mid-range model and GPT-OSS 120B (Workers AI) a economy one. The benchmark cannot separate their recall: Gemini 3.8 Flash minus GPT-OSS 120B (Workers AI) is +28 pts (95% interval +13 pts to +42 pts). - [Gemini 4 Argon vs GPT-OSS 120B (Workers AI)](https://codepanel.ai/benchmarks/compare/gemini-4-argon-vs-gpt-oss-120b-workers-ai): Reading a 26.2K-token pull request costs $0.0524 on Gemini 4 Argon, 5.7× the $0.0092 it costs on GPT-OSS 120B (Workers AI). Gemini 4 Argon is a mid-range model and GPT-OSS 120B (Workers AI) a economy one. In the benchmark Gemini 4 Argon caught 42 percentage points more of the known defects than GPT-OSS 120B (Workers AI) (95% interval +29 pts to +55 pts, a − b). - [GPT-5.6 Sol vs GPT-OSS 120B (Workers AI)](https://codepanel.ai/benchmarks/compare/gpt-5.6-sol-vs-gpt-oss-120b-workers-ai): Reading a 26.2K-token pull request costs $0.1048 on GPT-5.6 Sol, 11× the $0.0092 it costs on GPT-OSS 120B (Workers AI). GPT-5.6 Sol is a frontier model and GPT-OSS 120B (Workers AI) a economy one. In the benchmark GPT-5.6 Sol caught 32 percentage points more of the known defects than GPT-OSS 120B (Workers AI) (95% interval +18 pts to +46 pts, a − b). - [GPT-6 Luna vs GPT-OSS 120B (Workers AI)](https://codepanel.ai/benchmarks/compare/gpt-6-luna-vs-gpt-oss-120b-workers-ai): Reading a 26.2K-token pull request costs $0.0092 on GPT-OSS 120B (Workers AI), 3.5× the $0.0026 it costs on GPT-6 Luna. Both sit in the economy price tier, so the choice between them is about review quality more than budget. In the benchmark GPT-6 Luna caught 27 percentage points more of the known defects than GPT-OSS 120B (Workers AI) (95% interval +13 pts to +41 pts, a − b). - [Gemini 3.1 Pro vs GPT-OSS 120B (Workers AI)](https://codepanel.ai/benchmarks/compare/gemini-3.1-pro-vs-gpt-oss-120b-workers-ai): Reading a 26.2K-token pull request costs $0.0524 on Gemini 3.1 Pro, 5.7× the $0.0092 it costs on GPT-OSS 120B (Workers AI). Gemini 3.1 Pro is a mid-range model and GPT-OSS 120B (Workers AI) a economy one. In the benchmark Gemini 3.1 Pro caught 27 percentage points more of the known defects than GPT-OSS 120B (Workers AI) (95% interval +16 pts to +38 pts, a − b). - [GLM-5.3 (Workers AI) vs GPT-OSS 120B (Workers AI)](https://codepanel.ai/benchmarks/compare/glm-5.3-workers-ai-vs-gpt-oss-120b-workers-ai): Reading a 26.2K-token pull request costs $0.0367 on GLM-5.3 (Workers AI), 4.0× the $0.0092 it costs on GPT-OSS 120B (Workers AI). GLM-5.3 (Workers AI) is a mid-range model and GPT-OSS 120B (Workers AI) a economy one. The benchmark cannot separate their recall: GLM-5.3 (Workers AI) minus GPT-OSS 120B (Workers AI) is +21 pts (95% interval +10 pts to +32 pts). - [Claude Haiku 5.5 vs Kimi K2.7 Code (Workers AI)](https://codepanel.ai/benchmarks/compare/claude-haiku-5-5-vs-kimi-k2.7-code-workers-ai): Reading a 26.2K-token pull request costs $0.0249 on Kimi K2.7 Code (Workers AI), 9.5× the $0.0026 it costs on Claude Haiku 5.5. Claude Haiku 5.5 is a economy model and Kimi K2.7 Code (Workers AI) a mid-range one. The benchmark cannot separate their recall: Claude Haiku 5.5 minus Kimi K2.7 Code (Workers AI) is +26 pts (95% interval +11 pts to +41 pts). - [Claude Opus 5.5 vs Kimi K2.7 Code (Workers AI)](https://codepanel.ai/benchmarks/compare/claude-opus-5-5-vs-kimi-k2.7-code-workers-ai): Reading a 26.2K-token pull request costs $0.1048 on Claude Opus 5.5, 4.2× the $0.0249 it costs on Kimi K2.7 Code (Workers AI). Claude Opus 5.5 is a frontier model and Kimi K2.7 Code (Workers AI) a mid-range one. In the benchmark Claude Opus 5.5 caught 52 percentage points more of the known defects than Kimi K2.7 Code (Workers AI) (95% interval +41 pts to +63 pts, a − b). - [Gemini 3.8 Flash vs Kimi K2.7 Code (Workers AI)](https://codepanel.ai/benchmarks/compare/gemini-3.8-flash-vs-kimi-k2.7-code-workers-ai): Gemini 3.8 Flash reads a 26.2K-token pull request for $0.0197, 21% less than Kimi K2.7 Code (Workers AI) ($0.0249). From 2027-01-01, after a scheduled reprice, Kimi K2.7 Code (Workers AI) becomes the cheaper one. Both sit in the mid-range price tier, so the choice between them is about review quality more than budget. In the benchmark Gemini 3.8 Flash caught 30 percentage points more of the known defects than Kimi K2.7 Code (Workers AI) (95% interval +16 pts to +44 pts, a − b). - [GPT-6 Luna vs Kimi K2.7 Code (Workers AI)](https://codepanel.ai/benchmarks/compare/gpt-6-luna-vs-kimi-k2.7-code-workers-ai): Reading a 26.2K-token pull request costs $0.0249 on Kimi K2.7 Code (Workers AI), 9.5× the $0.0026 it costs on GPT-6 Luna. GPT-6 Luna is a economy model and Kimi K2.7 Code (Workers AI) a mid-range one. In the benchmark GPT-6 Luna caught 30 percentage points more of the known defects than Kimi K2.7 Code (Workers AI) (95% interval +14 pts to +44 pts, a − b). - [DeepSeek V4 Flash (Workers AI) vs GPT-OSS 120B (Workers AI)](https://codepanel.ai/benchmarks/compare/deepseek-v4-flash-workers-ai-vs-gpt-oss-120b-workers-ai): GPT-OSS 120B (Workers AI) reads a 26.2K-token pull request for $0.0092, 20% less than DeepSeek V4 Flash (Workers AI) ($0.0115). Both sit in the economy price tier, so the choice between them is about review quality more than budget. The benchmark cannot separate their recall: DeepSeek V4 Flash (Workers AI) minus GPT-OSS 120B (Workers AI) is +8 pts (95% interval −3 pts to +20 pts). - [Gemini 3.1 Pro vs Kimi K2.7 Code (Workers AI)](https://codepanel.ai/benchmarks/compare/gemini-3.1-pro-vs-kimi-k2.7-code-workers-ai): Reading a 26.2K-token pull request costs $0.0524 on Gemini 3.1 Pro, 2.1× the $0.0249 it costs on Kimi K2.7 Code (Workers AI). Both sit in the mid-range price tier, so the choice between them is about review quality more than budget. In the benchmark Gemini 3.1 Pro caught 30 percentage points more of the known defects than Kimi K2.7 Code (Workers AI) (95% interval +16 pts to +44 pts, a − b). - [GLM-5.3 Flash (Workers AI) vs Kimi K2.7 Code (Workers AI)](https://codepanel.ai/benchmarks/compare/glm-5.3-flash-workers-ai-vs-kimi-k2.7-code-workers-ai): Reading a 26.2K-token pull request costs $0.0249 on Kimi K2.7 Code (Workers AI), 6.3× the $0.0039 it costs on GLM-5.3 Flash (Workers AI). GLM-5.3 Flash (Workers AI) is a economy model and Kimi K2.7 Code (Workers AI) a mid-range one. In the benchmark GLM-5.3 Flash (Workers AI) caught 23 percentage points more of the known defects than Kimi K2.7 Code (Workers AI) (95% interval +11 pts to +35 pts, a − b). - [Claude Haiku 5.5 vs Gemma 4 26B A4B (Workers AI)](https://codepanel.ai/benchmarks/compare/claude-haiku-5-5-vs-gemma-4-26b-a4b-workers-ai): Claude Haiku 5.5 and Gemma 4 26B A4B (Workers AI) charge practically the same to read a 26.2K-token pull request ($0.0026). Above 100K tokens long-context rates apply, and Gemma 4 26B A4B (Workers AI) becomes the cheaper one. Both sit in the economy price tier, so the choice between them is about review quality more than budget. In the benchmark Claude Haiku 5.5 caught 38 percentage points more of the known defects than Gemma 4 26B A4B (Workers AI) (95% interval +22 pts to +53 pts, a − b). - [Claude Sonnet 5.5 vs Gemma 4 26B A4B (Workers AI)](https://codepanel.ai/benchmarks/compare/claude-sonnet-5-5-vs-gemma-4-26b-a4b-workers-ai): Reading a 26.2K-token pull request costs $0.0524 on Claude Sonnet 5.5, 20× the $0.0026 it costs on Gemma 4 26B A4B (Workers AI). Claude Sonnet 5.5 is a mid-range model and Gemma 4 26B A4B (Workers AI) a economy one. In the benchmark Claude Sonnet 5.5 caught 51 percentage points more of the known defects than Gemma 4 26B A4B (Workers AI) (95% interval +35 pts to +66 pts, a − b). - [Gemini 3.8 Flash vs Gemma 4 26B A4B (Workers AI)](https://codepanel.ai/benchmarks/compare/gemini-3.8-flash-vs-gemma-4-26b-a4b-workers-ai): Reading a 26.2K-token pull request costs $0.0197 on Gemini 3.8 Flash, 7.5× the $0.0026 it costs on Gemma 4 26B A4B (Workers AI). Gemini 3.8 Flash is a mid-range model and Gemma 4 26B A4B (Workers AI) a economy one. In the benchmark Gemini 3.8 Flash caught 42 percentage points more of the known defects than Gemma 4 26B A4B (Workers AI) (95% interval +25 pts to +58 pts, a − b). - [Gemma 4 26B A4B (Workers AI) vs GPT-6 Luna](https://codepanel.ai/benchmarks/compare/gemma-4-26b-a4b-workers-ai-vs-gpt-6-luna): Gemma 4 26B A4B (Workers AI) and GPT-6 Luna charge practically the same to read a 26.2K-token pull request ($0.0026). Above 272K tokens long-context rates apply, and Gemma 4 26B A4B (Workers AI) becomes the cheaper one. Both sit in the economy price tier, so the choice between them is about review quality more than budget. In the benchmark GPT-6 Luna caught 41 percentage points more of the known defects than Gemma 4 26B A4B (Workers AI) (95% interval −56 pts to −26 pts, a − b). - [DeepSeek V4 Pro (Workers AI) vs Kimi K2.7 Code (Workers AI)](https://codepanel.ai/benchmarks/compare/deepseek-v4-pro-workers-ai-vs-kimi-k2.7-code-workers-ai): Kimi K2.7 Code (Workers AI) reads a 26.2K-token pull request for $0.0249, 28% less than DeepSeek V4 Pro (Workers AI) ($0.0346). Both sit in the mid-range price tier, so the choice between them is about review quality more than budget. The benchmark cannot separate their recall: DeepSeek V4 Pro (Workers AI) minus Kimi K2.7 Code (Workers AI) is +16 pts (95% interval +8 pts to +24 pts). - [Claude Haiku 5.5 vs Gemini 3.5 Flash-Lite](https://codepanel.ai/benchmarks/compare/claude-haiku-5-5-vs-gemini-3.5-flash-lite): Reading a 26.2K-token pull request costs $0.0079 on Gemini 3.5 Flash-Lite, 3.0× the $0.0026 it costs on Claude Haiku 5.5. Above 100K tokens long-context rates apply, and Gemini 3.5 Flash-Lite becomes the cheaper one. Both sit in the economy price tier, so the choice between them is about review quality more than budget. In the benchmark Claude Haiku 5.5 caught 53 percentage points more of the known defects than Gemini 3.5 Flash-Lite (95% interval +37 pts to +67 pts, a − b). - [DeepSeek V4 Pro (Workers AI) vs Gemma 4 26B A4B (Workers AI)](https://codepanel.ai/benchmarks/compare/deepseek-v4-pro-workers-ai-vs-gemma-4-26b-a4b-workers-ai): Reading a 26.2K-token pull request costs $0.0346 on DeepSeek V4 Pro (Workers AI), 13× the $0.0026 it costs on Gemma 4 26B A4B (Workers AI). DeepSeek V4 Pro (Workers AI) is a mid-range model and Gemma 4 26B A4B (Workers AI) a economy one. In the benchmark DeepSeek V4 Pro (Workers AI) caught 28 percentage points more of the known defects than Gemma 4 26B A4B (Workers AI) (95% interval +14 pts to +42 pts, a − b). - [Gemini 3.5 Flash-Lite vs GPT-6 Luna](https://codepanel.ai/benchmarks/compare/gemini-3.5-flash-lite-vs-gpt-6-luna): Reading a 26.2K-token pull request costs $0.0079 on Gemini 3.5 Flash-Lite, 3.0× the $0.0026 it costs on GPT-6 Luna. Both sit in the economy price tier, so the choice between them is about review quality more than budget. In the benchmark GPT-6 Luna caught 56 percentage points more of the known defects than Gemini 3.5 Flash-Lite (95% interval −71 pts to −40 pts, a − b). - [Gemini 3.5 Flash-Lite vs GLM-5.3 Flash (Workers AI)](https://codepanel.ai/benchmarks/compare/gemini-3.5-flash-lite-vs-glm-5.3-flash-workers-ai): Reading a 26.2K-token pull request costs $0.0079 on Gemini 3.5 Flash-Lite, 2.0× the $0.0039 it costs on GLM-5.3 Flash (Workers AI). Both sit in the economy price tier, so the choice between them is about review quality more than budget. In the benchmark GLM-5.3 Flash (Workers AI) caught 50 percentage points more of the known defects than Gemini 3.5 Flash-Lite (95% interval −61 pts to −38 pts, a − b). - [Gemini 3.5 Flash-Lite vs Gemma 4 26B A4B (Workers AI)](https://codepanel.ai/benchmarks/compare/gemini-3.5-flash-lite-vs-gemma-4-26b-a4b-workers-ai): Reading a 26.2K-token pull request costs $0.0079 on Gemini 3.5 Flash-Lite, 3.0× the $0.0026 it costs on Gemma 4 26B A4B (Workers AI). Both sit in the economy price tier, so the choice between them is about review quality more than budget. The benchmark cannot separate their recall: Gemini 3.5 Flash-Lite minus Gemma 4 26B A4B (Workers AI) is −15 pts (95% interval −27 pts to −3 pts). ## Models - [AI models](https://codepanel.ai/models): CodePanel can run 76 AI models from 15 makers as pull-request reviewers. A typical review call costs from $0.0027 on GPT-OSS 20B (Groq) to $0.3829 on GPT-6 Astra at list price. Each maker's page lists its models' rates and what a review costs on each. - [LLM API pricing](https://codepanel.ai/models/pricing): CodePanel prices 76 current models from 15 makers at their list rates: $0.0341 to $10.00 per 1M input tokens and $0.30 to $50.00 per 1M output tokens. One review call (26.2K tokens in, 2.4K out) costs from $0.0027 on GPT-OSS 20B (Groq) to $0.3829 on Claude Fable 5.1. With a token cost calculator; the same prices as JSON at https://codepanel.ai/models/pricing.json. - [Anthropic Claude API pricing](https://codepanel.ai/models/anthropic): The Anthropic Claude API charges $0.10 to $10.00 per 1M input tokens and $0.50 to $50.00 per 1M output tokens across the 12 current models CodePanel offers. One review call (26.2K tokens in, 2.4K out) costs from $0.0038 on Claude Haiku 5.5 to $0.3829 on Claude Fable 5.1. - [OpenAI API pricing](https://codepanel.ai/models/openai): The OpenAI API charges $0.10 to $10.00 per 1M input tokens and $0.50 to $50.00 per 1M output tokens across the 12 current models CodePanel offers. One review call (26.2K tokens in, 2.4K out) costs from $0.0038 on GPT-6 Luna to $0.3829 on GPT-6 Astra. - [Google Gemini API pricing](https://codepanel.ai/models/google): The Google Gemini API charges $0.10 to $2.00 per 1M input tokens and $0.40 to $12.00 per 1M output tokens across the 9 current models CodePanel offers. One review call (26.2K tokens in, 2.4K out) costs from $0.0036 on Gemini 2.5 Flash-Lite to $0.0814 on Gemini 3.1 Pro. - [xAI Grok API pricing](https://codepanel.ai/models/xai): The xAI Grok API charges $1.00 to $2.00 per 1M input tokens and $2.00 to $6.00 per 1M output tokens across the 6 current models CodePanel offers. One review call (26.2K tokens in, 2.4K out) costs from $0.031 on Grok Build 0.1 to $0.0669 on Grok 4.7. - [DeepSeek API pricing](https://codepanel.ai/models/deepseek): The DeepSeek API charges $0.30 to $1.32 per 1M input tokens and $1.20 to $3.96 per 1M output tokens across the 2 current models CodePanel offers. One review call (26.2K tokens in, 2.4K out) costs from $0.0108 on DeepSeek Flash to $0.0442 on DeepSeek V4 Pro. - [Moonshot Kimi API pricing](https://codepanel.ai/models/moonshot): The Moonshot Kimi API charges $0.95 to $3.00 per 1M input tokens and $4.00 to $15.00 per 1M output tokens across the 4 current models CodePanel offers. One review call (26.2K tokens in, 2.4K out) costs from $0.0346 on Kimi K2.6 to $0.1149 on Kimi K3. - [Mistral API pricing](https://codepanel.ai/models/mistral): The Mistral API charges $0.30 to $1.50 per 1M input tokens and $0.90 to $7.50 per 1M output tokens across the 2 current models CodePanel offers. One review call (26.2K tokens in, 2.4K out) costs from $0.01 on Codestral 25.08 to $0.0574 on Mistral Medium 3.5. - [Open-weight models on Groq](https://codepanel.ai/models/groq): Groq serves 3 open-weight models from other makers that CodePanel can run: $0.075 to $0.80 per 1M input tokens and $0.30 to $4.00 per 1M output tokens. One review call (26.2K tokens in, 2.4K out) costs from $0.0027 on GPT-OSS 20B (Groq) to $0.0306 on Qwen3.8 27B (Groq). - [Z.ai models on Cloudflare Workers AI and OpenRouter](https://codepanel.ai/models/zai): CodePanel offers 5 Z.ai models, priced on Cloudflare Workers AI and OpenRouter: $0.0341 to $1.40 per 1M input tokens and $0.50 to $7.00 per 1M output tokens. One review call (26.2K tokens in, 2.4K out) costs from $0.0051 on GLM-5.3 Flash (Workers AI) to $0.0473 on GLM-5.3 (Workers AI). - [Qwen models on OpenRouter and Cloudflare Workers AI](https://codepanel.ai/models/qwen): CodePanel offers 3 Qwen models, priced on OpenRouter and Cloudflare Workers AI: $0.45 to $2.00 per 1M input tokens and $3.20 to $6.00 per 1M output tokens. One review call (26.2K tokens in, 2.4K out) costs from $0.0195 on Qwen3.8 27B (Workers AI) to $0.0669 on Qwen3.8 Max (OpenRouter). - [Claude Fable 5.1: benchmark, pricing and specs](https://codepanel.ai/models/anthropic/claude-fable-5-1): Claude Fable 5.1 is an Anthropic model. It costs $10.00 per 1M input tokens and $50.00 per 1M output tokens, about $0.3829 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is about the same as Claude Fable 5 and about the same as GPT-6 Astra, the closest-priced model from another provider. - [Claude Opus 5.5: benchmark, pricing and specs](https://codepanel.ai/models/anthropic/claude-opus-5-5): Claude Opus 5.5 is an Anthropic model. In CodePanel's benchmark it caught 90% [81%–97%] of known defects and raised 1.7 false alarms per clean pull request, at a measured $0.1525 per review and $0.1921 per bug found. It costs $4.00 per 1M input tokens and $20.00 per 1M output tokens, about $0.1532 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 20% less than Claude Opus 5 and about the same as GPT-5.6 Sol, the closest-priced model from another provider. - [Claude Opus 5: benchmark, pricing and specs](https://codepanel.ai/models/anthropic/claude-opus-5): Claude Opus 5 is an Anthropic model. It costs $5.00 per 1M input tokens and $25.00 per 1M output tokens, about $0.1915 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 25% more than Claude Opus 5.5 and 6% less than GPT-5.5, the closest-priced model from another provider. - [Claude Sonnet 5.5: benchmark, pricing and specs](https://codepanel.ai/models/anthropic/claude-sonnet-5-5): Claude Sonnet 5.5 is an Anthropic model. In CodePanel's benchmark it caught 77% [64%–89%] of known defects and raised 2.0 false alarms per clean pull request, at a measured $0.0571 per review and $0.083 per bug found. It costs $2.00 per 1M input tokens and $10.00 per 1M output tokens, about $0.0766 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is about the same as Claude Sonnet 5 and about the same as GPT-6.1 Sol, the closest-priced model from another provider. - [Claude Sonnet 5: benchmark, pricing and specs](https://codepanel.ai/models/anthropic/claude-sonnet-5): Claude Sonnet 5 is an Anthropic model. It costs $2.00 per 1M input tokens and $10.00 per 1M output tokens, about $0.0766 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is about the same as Claude Sonnet 5.5 and about the same as GPT-6.1 Sol, the closest-priced model from another provider. - [Claude Sonnet 4.6: benchmark, pricing and specs](https://codepanel.ai/models/anthropic/claude-sonnet-4-6): Claude Sonnet 4.6 is an Anthropic model. It costs $3.00 per 1M input tokens and $15.00 per 1M output tokens, about $0.1149 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 50% more than Claude Sonnet 5.5 and about the same as Kimi K3, the closest-priced model from another provider. - [Claude Haiku 5.5: benchmark, pricing and specs](https://codepanel.ai/models/anthropic/claude-haiku-5-5): Claude Haiku 5.5 is an Anthropic model. In CodePanel's benchmark it caught 64% [50%–78%] of known defects and raised 1.4 false alarms per clean pull request, at a measured $0.0038 per review and $0.0064 per bug found. It costs $0.10 per 1M input tokens and $0.50 per 1M output tokens, about $0.0038 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 90% less than Claude Haiku 4.5 and about the same as GPT-6 Luna, the closest-priced model from another provider. - [Claude Haiku 4.5: benchmark, pricing and specs](https://codepanel.ai/models/anthropic/claude-haiku-4-5): Claude Haiku 4.5 is an Anthropic model. It costs $1.00 per 1M input tokens and $5.00 per 1M output tokens, about $0.0383 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 10.0× the cost of Claude Haiku 5.5 and about the same as Grok 4.3, the closest-priced model from another provider. - [GPT-6 Astra: benchmark, pricing and specs](https://codepanel.ai/models/openai/gpt-6-astra): GPT-6 Astra is an OpenAI model. It costs $10.00 per 1M input tokens and $50.00 per 1M output tokens, about $0.3829 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is about the same as Claude Fable 5.1, the closest-priced model from another provider. - [GPT-6.1 Sol: benchmark, pricing and specs](https://codepanel.ai/models/openai/gpt-6.1-sol): GPT-6.1 Sol is an OpenAI model. In CodePanel's benchmark it caught 89% [80%–97%] of known defects and raised 0.9 false alarms per clean pull request, at a measured $0.0401 per review and $0.0514 per bug found. It costs $2.00 per 1M input tokens and $10.00 per 1M output tokens, about $0.0766 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is about the same as GPT-6 Sol and about the same as Claude Sonnet 5.5, the closest-priced model from another provider. - [GPT-6 Sol: benchmark, pricing and specs](https://codepanel.ai/models/openai/gpt-6-sol): GPT-6 Sol is an OpenAI model. It costs $2.00 per 1M input tokens and $10.00 per 1M output tokens, about $0.0766 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is about the same as GPT-6.1 Sol and about the same as Claude Sonnet 5.5, the closest-priced model from another provider. - [GPT-5.6 Sol: benchmark, pricing and specs](https://codepanel.ai/models/openai/gpt-5.6-sol): GPT-5.6 Sol is an OpenAI model. In CodePanel's benchmark it caught 72% [59%–85%] of known defects and raised 1.2 false alarms per clean pull request, at a measured $0.1127 per review and $0.1732 per bug found. It costs $4.00 per 1M input tokens and $20.00 per 1M output tokens, about $0.1532 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 2.0× the cost of GPT-6.1 Sol and about the same as Claude Opus 5.5, the closest-priced model from another provider. - [GPT-5.6 Terra: benchmark, pricing and specs](https://codepanel.ai/models/openai/gpt-5.6-terra): GPT-5.6 Terra is an OpenAI model. It costs $2.00 per 1M input tokens and $12.00 per 1M output tokens, about $0.0814 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is about the same as Gemini 3.1 Pro, the closest-priced model from another provider. - [GPT-6 Luna: benchmark, pricing and specs](https://codepanel.ai/models/openai/gpt-6-luna): GPT-6 Luna is an OpenAI model. In CodePanel's benchmark it caught 67% [53%–80%] of known defects and raised 0.7 false alarms per clean pull request, at a measured $0.0028 per review and $0.0046 per bug found. It costs $0.10 per 1M input tokens and $0.50 per 1M output tokens, about $0.0038 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 53% less than GPT-5.6 Luna and about the same as Claude Haiku 5.5, the closest-priced model from another provider. - [GPT-5.6 Luna: benchmark, pricing and specs](https://codepanel.ai/models/openai/gpt-5.6-luna): GPT-5.6 Luna is an OpenAI model. It costs $0.20 per 1M input tokens and $1.20 per 1M output tokens, about $0.0081 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 2.1× the cost of GPT-6 Luna and 19% less than Codestral 25.08, the closest-priced model from another provider. - [GPT-5.5 Pro: benchmark, pricing and specs](https://codepanel.ai/models/openai/gpt-5.5-pro): GPT-5.5 Pro is an OpenAI model. It costs $30.00 per 1M input tokens and $180.00 per 1M output tokens, about $1.2213 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is about the same as GPT-5.4 Pro and 3.2× the cost of Claude Fable 5.1, the closest-priced model from another provider. - [GPT-5.5: benchmark, pricing and specs](https://codepanel.ai/models/openai/gpt-5.5): GPT-5.5 is an OpenAI model. It costs $5.00 per 1M input tokens and $30.00 per 1M output tokens, about $0.2035 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 2.0× the cost of GPT-5.4 and 6% more than Claude Opus 5, the closest-priced model from another provider. - [GPT-5.4: benchmark, pricing and specs](https://codepanel.ai/models/openai/gpt-5.4): GPT-5.4 is an OpenAI model. It costs $2.50 per 1M input tokens and $15.00 per 1M output tokens, about $0.1018 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 50% less than GPT-5.5 and 11% less than Claude Sonnet 4.6, the closest-priced model from another provider. - [GPT-5.4 Mini: benchmark, pricing and specs](https://codepanel.ai/models/openai/gpt-5.4-mini): GPT-5.4 Mini is an OpenAI model. It costs $0.75 per 1M input tokens and $4.50 per 1M output tokens, about $0.0305 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is about the same as Qwen3.8 27B (Groq), the closest-priced model from another provider. - [GPT-5.4 Nano: benchmark, pricing and specs](https://codepanel.ai/models/openai/gpt-5.4-nano): GPT-5.4 Nano is an OpenAI model. It costs $0.20 per 1M input tokens and $1.25 per 1M output tokens, about $0.0083 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 18% less than Codestral 25.08, the closest-priced model from another provider. - [GPT-5.3 Codex: benchmark, pricing and specs](https://codepanel.ai/models/openai/gpt-5.3-codex): GPT-5.3 Codex is an OpenAI model. It costs $1.75 per 1M input tokens and $14.00 per 1M output tokens, about $0.0797 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is about the same as Gemini 3.1 Pro, the closest-priced model from another provider. - [GPT-OSS 120B (Workers AI): benchmark, pricing and specs](https://codepanel.ai/models/openai/gpt-oss-120b-workers-ai): GPT-OSS 120B (Workers AI) is an OpenAI model served by Cloudflare Workers AI. In CodePanel's benchmark it caught 40% [29%–52%] of known defects and raised 2.5 false alarms per clean pull request, at a measured $0.0075 per review and $0.0189 per bug found. It costs $0.35 per 1M input tokens and $0.75 per 1M output tokens, about $0.011 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is about the same as DeepSeek Flash, the closest-priced model from another provider. - [Gemini 4 Argon: benchmark, pricing and specs](https://codepanel.ai/models/google/gemini-4-argon): Gemini 4 Argon is a Google model. In CodePanel's benchmark it caught 82% [71%–92%] of known defects and raised 1.1 false alarms per clean pull request, at a measured $0.0527 per review and $0.0706 per bug found. It costs $2.00 per 1M input tokens and $10.00 per 1M output tokens, about $0.0766 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is about the same as Claude Sonnet 5.5, the closest-priced model from another provider. - [Gemini 3.1 Pro: benchmark, pricing and specs](https://codepanel.ai/models/google/gemini-3.1-pro): Gemini 3.1 Pro is a Google model. In CodePanel's benchmark it caught 67% [54%–80%] of known defects and raised 0.8 false alarms per clean pull request, at a measured $0.2499 per review and $0.4134 per bug found. It costs $2.00 per 1M input tokens and $12.00 per 1M output tokens, about $0.0814 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 43% more than Gemini 2.5 Pro and about the same as GPT-5.6 Terra, the closest-priced model from another provider. - [Gemini 2.5 Pro: benchmark, pricing and specs](https://codepanel.ai/models/google/gemini-2.5-pro): Gemini 2.5 Pro is a Google model. It costs $1.25 per 1M input tokens and $10.00 per 1M output tokens, about $0.0569 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 30% less than Gemini 3.1 Pro and about the same as GPT-5, the closest-priced model from another provider. - [Gemini 3.8 Flash: benchmark, pricing and specs](https://codepanel.ai/models/google/gemini-3.8-flash): Gemini 3.8 Flash is a Google model. In CodePanel's benchmark it caught 68% [54%–81%] of known defects and raised 1.0 false alarms per clean pull request, at a measured $0.0862 per review and $0.1166 per bug found. It costs $0.75 per 1M input tokens and $3.75 per 1M output tokens, about $0.0287 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is about the same as Gemini 3.6 Flash and 6% less than GPT-5.4 Mini, the closest-priced model from another provider. - [Gemini 3.6 Flash: benchmark, pricing and specs](https://codepanel.ai/models/google/gemini-3.6-flash): Gemini 3.6 Flash is a Google model. It costs $0.75 per 1M input tokens and $3.75 per 1M output tokens, about $0.0287 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is about the same as Gemini 3.8 Flash and 6% less than GPT-5.4 Mini, the closest-priced model from another provider. - [Gemini 3.5 Flash: benchmark, pricing and specs](https://codepanel.ai/models/google/gemini-3.5-flash): Gemini 3.5 Flash is a Google model, now retired by its provider. It costs $1.50 per 1M input tokens and $9.00 per 1M output tokens, about $0.0611 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 2.1× the cost of Gemini 3.8 Flash and 6% more than Mistral Medium 3.5, the closest-priced model from another provider. - [Gemini 2.5 Flash: benchmark, pricing and specs](https://codepanel.ai/models/google/gemini-2.5-flash): Gemini 2.5 Flash is a Google model. It costs $0.30 per 1M input tokens and $2.50 per 1M output tokens, about $0.0139 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 52% less than Gemini 3.8 Flash and about the same as MiMo-V2.6-Pro (OpenRouter), the closest-priced model from another provider. - [Gemini 3.5 Flash-Lite: benchmark, pricing and specs](https://codepanel.ai/models/google/gemini-3.5-flash-lite): Gemini 3.5 Flash-Lite is a Google model. In CodePanel's benchmark it caught 11% [3%–21%] of known defects and raised 0.4 false alarms per clean pull request, at a measured $0.0059 per review and $0.0528 per bug found. It costs $0.30 per 1M input tokens and $2.50 per 1M output tokens, about $0.0139 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 37% more than Gemini 3.1 Flash-Lite and about the same as MiMo-V2.6-Pro (OpenRouter), the closest-priced model from another provider. - [Gemini 3.1 Flash-Lite: benchmark, pricing and specs](https://codepanel.ai/models/google/gemini-3.1-flash-lite): Gemini 3.1 Flash-Lite is a Google model. It costs $0.25 per 1M input tokens and $1.50 per 1M output tokens, about $0.0102 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 27% less than Gemini 3.5 Flash-Lite and about the same as Codestral 25.08, the closest-priced model from another provider. - [Gemini 2.5 Flash-Lite: benchmark, pricing and specs](https://codepanel.ai/models/google/gemini-2.5-flash-lite): Gemini 2.5 Flash-Lite is a Google model. It costs $0.10 per 1M input tokens and $0.40 per 1M output tokens, about $0.0036 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 74% less than Gemini 3.5 Flash-Lite and 6% less than Claude Haiku 5.5, the closest-priced model from another provider. - [Gemma 4 26B A4B (Workers AI): benchmark, pricing and specs](https://codepanel.ai/models/google/gemma-4-26b-a4b-workers-ai): Gemma 4 26B A4B (Workers AI) is a Google model served by Cloudflare Workers AI. In CodePanel's benchmark it caught 26% [15%–38%] of known defects and raised 0.4 false alarms per clean pull request, at a measured $0.0048 per review and $0.0209 per bug found. It costs $0.10 per 1M input tokens and $0.30 per 1M output tokens, about $0.0033 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 13% less than Claude Haiku 5.5, the closest-priced model from another provider. - [Grok 4.7: benchmark, pricing and specs](https://codepanel.ai/models/xai/grok-4.7): Grok 4.7 is an xAI model. It costs $2.00 per 1M input tokens and $6.00 per 1M output tokens, about $0.0669 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is about the same as Grok 4.6 and about the same as Qwen3.8 Max (OpenRouter), the closest-priced model from another provider. - [Grok 4.3: benchmark, pricing and specs](https://codepanel.ai/models/xai/grok-4.3): Grok 4.3 is an xAI model. It costs $1.25 per 1M input tokens and $2.50 per 1M output tokens, about $0.0388 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 42% less than Grok 4.7 and about the same as Claude Haiku 4.5, the closest-priced model from another provider. - [Grok Build 0.1: benchmark, pricing and specs](https://codepanel.ai/models/xai/grok-build-0.1): Grok Build 0.1 is an xAI model. It costs $1.00 per 1M input tokens and $2.00 per 1M output tokens, about $0.031 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is about the same as Qwen3.8 27B (Groq), the closest-priced model from another provider. - [DeepSeek V4 Pro: benchmark, pricing and specs](https://codepanel.ai/models/deepseek/deepseek-v4-pro): DeepSeek V4 Pro is a DeepSeek model. It costs $1.32 per 1M input tokens and $3.96 per 1M output tokens, about $0.0442 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is about the same as DeepSeek V4 Pro (Workers AI) and about the same as Muse Spark 1.2, the closest-priced model from another provider. - [DeepSeek Flash: benchmark, pricing and specs](https://codepanel.ai/models/deepseek/deepseek-flash): DeepSeek Flash is a DeepSeek model. It costs $0.30 per 1M input tokens and $1.20 per 1M output tokens, about $0.0108 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 27% less than DeepSeek V4 Flash (Workers AI) and about the same as GPT-OSS 120B (Workers AI), the closest-priced model from another provider. - [DeepSeek V4 Pro (Workers AI): benchmark, pricing and specs](https://codepanel.ai/models/deepseek/deepseek-v4-pro-workers-ai): DeepSeek V4 Pro (Workers AI) is a DeepSeek model served by Cloudflare Workers AI. In CodePanel's benchmark it caught 54% [43%–65%] of known defects and raised 2.9 false alarms per clean pull request, at a measured $0.0643 per review and $0.0872 per bug found. It costs $1.32 per 1M input tokens and $3.96 per 1M output tokens, about $0.0442 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is about the same as DeepSeek V4 Pro and about the same as Muse Spark 1.2, the closest-priced model from another provider. - [DeepSeek V4 Flash (Workers AI): benchmark, pricing and specs](https://codepanel.ai/models/deepseek/deepseek-v4-flash-workers-ai): DeepSeek V4 Flash (Workers AI) is a DeepSeek model served by Cloudflare Workers AI. In CodePanel's benchmark it caught 49% [37%–60%] of known defects and raised 1.3 false alarms per clean pull request, at a measured $0.0178 per review and $0.0309 per bug found. It costs $0.44 per 1M input tokens and $1.32 per 1M output tokens, about $0.0147 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 37% more than DeepSeek Flash and about the same as Inkling Small, the closest-priced model from another provider. - [Kimi K3: benchmark, pricing and specs](https://codepanel.ai/models/moonshot/kimi-k3): Kimi K3 is a Moonshot model. It costs $3.00 per 1M input tokens and $15.00 per 1M output tokens, about $0.1149 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 3.3× the cost of Kimi K2.6 and about the same as Claude Sonnet 4.6, the closest-priced model from another provider. - [Kimi K2.6: benchmark, pricing and specs](https://codepanel.ai/models/moonshot/kimi-k2.6): Kimi K2.6 is a Moonshot model. It costs $0.95 per 1M input tokens and $4.00 per 1M output tokens, about $0.0346 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 70% less than Kimi K3 and about the same as GLM-5, the closest-priced model from another provider. - [Kimi K2.7 Code Highspeed: benchmark, pricing and specs](https://codepanel.ai/models/moonshot/kimi-k2.7-code-highspeed): Kimi K2.7 Code Highspeed is a Moonshot model. It costs $1.90 per 1M input tokens and $8.00 per 1M output tokens, about $0.0691 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 2.0× the cost of Kimi K2.7 Code and about the same as Grok 4.7, the closest-priced model from another provider. - [Kimi K2.7 Code: benchmark, pricing and specs](https://codepanel.ai/models/moonshot/kimi-k2.7-code): Kimi K2.7 Code is a Moonshot model. It costs $0.95 per 1M input tokens and $4.00 per 1M output tokens, about $0.0346 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 50% less than Kimi K2.7 Code Highspeed and about the same as GLM-5, the closest-priced model from another provider. - [Kimi K2.7 Code (Workers AI): benchmark, pricing and specs](https://codepanel.ai/models/moonshot/kimi-k2.7-code-workers-ai): Kimi K2.7 Code (Workers AI) is a Moonshot model served by Cloudflare Workers AI. In CodePanel's benchmark it caught 38% [28%–47%] of known defects and raised 2.6 false alarms per clean pull request, at a measured $0.0567 per review and $0.0805 per bug found. It costs $0.95 per 1M input tokens and $4.00 per 1M output tokens, about $0.0346 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 50% less than Kimi K2.7 Code Highspeed and about the same as GLM-5, the closest-priced model from another provider. - [Mistral Medium 3.5: benchmark, pricing and specs](https://codepanel.ai/models/mistral/mistral-medium-3-5): Mistral Medium 3.5 is a Mistral model. It costs $1.50 per 1M input tokens and $7.50 per 1M output tokens, about $0.0574 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is about the same as Gemini 2.5 Pro, the closest-priced model from another provider. - [Codestral 25.08: benchmark, pricing and specs](https://codepanel.ai/models/mistral/codestral-2508): Codestral 25.08 is a Mistral model. It costs $0.30 per 1M input tokens and $0.90 per 1M output tokens, about $0.01 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is about the same as Gemini 3.1 Flash-Lite, the closest-priced model from another provider. - [GPT-OSS 120B (Groq): benchmark, pricing and specs](https://codepanel.ai/models/groq/groq-gpt-oss-120b): GPT-OSS 120B (Groq) is an OpenAI model served by Groq. It costs $0.15 per 1M input tokens and $0.60 per 1M output tokens, about $0.0054 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 2.0× the cost of GPT-OSS 20B (Groq) and about the same as GLM-5.3 Flash (Workers AI), the closest-priced model from another provider. - [GPT-OSS 20B (Groq): benchmark, pricing and specs](https://codepanel.ai/models/groq/groq-gpt-oss-20b): GPT-OSS 20B (Groq) is an OpenAI model served by Groq. It costs $0.075 per 1M input tokens and $0.30 per 1M output tokens, about $0.0027 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 50% less than GPT-OSS 120B (Groq) and 20% less than Gemma 4 26B A4B (Workers AI), the closest-priced model from another provider. - [Qwen3.8 27B (Groq): benchmark, pricing and specs](https://codepanel.ai/models/groq/groq-qwen3.8-27b): Qwen3.8 27B (Groq) is a Qwen model served by Groq. It costs $0.80 per 1M input tokens and $4.00 per 1M output tokens, about $0.0306 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is about the same as GPT-5.4 Mini, the closest-priced model from another provider. - [GLM-5.3 (Workers AI): benchmark, pricing and specs](https://codepanel.ai/models/zai/glm-5.3-workers-ai): GLM-5.3 (Workers AI) is a Z.ai model served by Cloudflare Workers AI. In CodePanel's benchmark it caught 61% [50%–72%] of known defects and raised 1.9 false alarms per clean pull request, at a measured $0.0311 per review and $0.0559 per bug found. It costs $1.40 per 1M input tokens and $4.40 per 1M output tokens, about $0.0473 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 2.7× the cost of GLM-5.2 and about the same as Mistral Large 4, the closest-priced model from another provider. - [GLM-5.3 Flash (Workers AI): benchmark, pricing and specs](https://codepanel.ai/models/zai/glm-5.3-flash-workers-ai): GLM-5.3 Flash (Workers AI) is a Z.ai model served by Cloudflare Workers AI. In CodePanel's benchmark it caught 61% [50%–72%] of known defects and raised 1.8 false alarms per clean pull request, at a measured $0.0035 per review and $0.0066 per bug found. It costs $0.15 per 1M input tokens and $0.50 per 1M output tokens, about $0.0051 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is about the same as GPT-OSS 120B (Groq), the closest-priced model from another provider. - [Qwen3.8 Max (OpenRouter): benchmark, pricing and specs](https://codepanel.ai/models/qwen/qwen3.8-max-openrouter): Qwen3.8 Max (OpenRouter) is a Qwen model served by OpenRouter. It costs $2.00 per 1M input tokens and $6.00 per 1M output tokens, about $0.0669 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 36% more than Qwen3.7 Max and about the same as Grok 4.7, the closest-priced model from another provider. - [Qwen3.8 27B (Workers AI): benchmark, pricing and specs](https://codepanel.ai/models/qwen/qwen3.8-27b-workers-ai): Qwen3.8 27B (Workers AI) is a Qwen model served by Cloudflare Workers AI. It costs $0.45 per 1M input tokens and $3.20 per 1M output tokens, about $0.0195 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is about the same as GLM-4.7, the closest-priced model from another provider. - [Nemotron 3 Super (Workers AI): benchmark, pricing and specs](https://codepanel.ai/models/nvidia/nemotron-3-super-workers-ai): Nemotron 3 Super (Workers AI) is an NVIDIA model served by Cloudflare Workers AI. It costs $0.50 per 1M input tokens and $1.50 per 1M output tokens, about $0.0167 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is about the same as Mistral Large 3, the closest-priced model from another provider. - [Step 5 Preview (OpenRouter): benchmark, pricing and specs](https://codepanel.ai/models/stepfun/step-5-preview-openrouter): Step 5 Preview (OpenRouter) is a StepFun model served by OpenRouter. It costs $1.00 per 1M input tokens and $2.70 per 1M output tokens, about $0.0327 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is about the same as GLM-5, the closest-priced model from another provider. - [MiMo-V2.6-Pro (OpenRouter): benchmark, pricing and specs](https://codepanel.ai/models/xiaomi/mimo-v2.6-pro-openrouter): MiMo-V2.6-Pro (OpenRouter) is a Xiaomi model served by OpenRouter. It costs $0.435 per 1M input tokens and $0.87 per 1M output tokens, about $0.0135 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 90% less than MiMo-V2.6-Pro UltraSpeed (OpenRouter) and about the same as Gemini 2.5 Flash, the closest-priced model from another provider. - [MiMo-V2.6-Pro UltraSpeed (OpenRouter): benchmark, pricing and specs](https://codepanel.ai/models/xiaomi/mimo-v2.6-pro-ultraspeed-openrouter): MiMo-V2.6-Pro UltraSpeed (OpenRouter) is a Xiaomi model served by OpenRouter. It costs $4.35 per 1M input tokens and $8.70 per 1M output tokens, about $0.135 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 10× the cost of MiMo-V2.6-Pro (OpenRouter) and 18% more than Claude Sonnet 4.6, the closest-priced model from another provider. - [Muse Spark 1.2: benchmark, pricing and specs](https://codepanel.ai/models/meta/muse-spark-1.2): Muse Spark 1.2 is a Meta model served by OpenRouter. It costs $1.25 per 1M input tokens and $4.25 per 1M output tokens, about $0.043 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is about the same as Muse Spark 1.1 and about the same as DeepSeek V4 Pro, the closest-priced model from another provider. - [Mistral Large 4: benchmark, pricing and specs](https://codepanel.ai/models/mistral/mistral-large-4-0): Mistral Large 4 is a Mistral model served by OpenRouter. It costs $1.36 per 1M input tokens and $4.18 per 1M output tokens, about $0.0457 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 2.7× the cost of Mistral Large 3 and about the same as DeepSeek V4 Pro, the closest-priced model from another provider. - [Mistral Large 3: benchmark, pricing and specs](https://codepanel.ai/models/mistral/mistral-large-2512): Mistral Large 3 is a Mistral model served by OpenRouter. It costs $0.50 per 1M input tokens and $1.50 per 1M output tokens, about $0.0167 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 63% less than Mistral Large 4 and about the same as Nemotron 3 Super (Workers AI), the closest-priced model from another provider. - [GPT-5.2: benchmark, pricing and specs](https://codepanel.ai/models/openai/gpt-5.2): GPT-5.2 is an OpenAI model served by OpenRouter. It costs $1.75 per 1M input tokens and $14.00 per 1M output tokens, about $0.0797 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 61% less than GPT-5.5 and about the same as Gemini 3.1 Pro, the closest-priced model from another provider. - [GPT-5: benchmark, pricing and specs](https://codepanel.ai/models/openai/gpt-5): GPT-5 is an OpenAI model served by OpenRouter. It costs $1.25 per 1M input tokens and $10.00 per 1M output tokens, about $0.0569 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 72% less than GPT-5.5 and about the same as Gemini 2.5 Pro, the closest-priced model from another provider. - [Qwen3.7 Max: benchmark, pricing and specs](https://codepanel.ai/models/qwen/qwen3.7-max): Qwen3.7 Max is a Qwen model served by OpenRouter. It costs $1.475 per 1M input tokens and $4.425 per 1M output tokens, about $0.0494 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 26% less than Qwen3.8 Max (OpenRouter) and about the same as GLM-5.3 (Workers AI), the closest-priced model from another provider. - [Inkling: benchmark, pricing and specs](https://codepanel.ai/models/thinkingmachines/inkling): Inkling is a Thinking Machines model served by OpenRouter. It costs $1.00 per 1M input tokens and $4.05 per 1M output tokens, about $0.036 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is about the same as Kimi K2.6, the closest-priced model from another provider. - [Inkling Small: benchmark, pricing and specs](https://codepanel.ai/models/thinkingmachines/inkling-small): Inkling Small is a Thinking Machines model served by OpenRouter. It costs $0.45 per 1M input tokens and $1.20 per 1M output tokens, about $0.0147 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is about the same as DeepSeek V4 Flash (Workers AI), the closest-priced model from another provider. - [GLM-5.2: benchmark, pricing and specs](https://codepanel.ai/models/zai/glm-5.2): GLM-5.2 is a Z.ai model served by OpenRouter. It costs $0.0341 per 1M input tokens and $7.00 per 1M output tokens, about $0.0178 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 62% less than GLM-5.3 (Workers AI) and 6% more than Nemotron 3 Super (Workers AI), the closest-priced model from another provider. - [GLM-5: benchmark, pricing and specs](https://codepanel.ai/models/zai/glm-5): GLM-5 is a Z.ai model served by OpenRouter. It costs $1.00 per 1M input tokens and $3.20 per 1M output tokens, about $0.0339 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 28% less than GLM-5.3 (Workers AI) and about the same as Kimi K2.6, the closest-priced model from another provider. - [GLM-4.7: benchmark, pricing and specs](https://codepanel.ai/models/zai/glm-4.7): GLM-4.7 is a Z.ai model served by OpenRouter. It costs $0.54 per 1M input tokens and $1.98 per 1M output tokens, about $0.0189 for a typical review call (26.2K tokens in, 2.4K out). Per review call that is 60% less than GLM-5.3 (Workers AI) and about the same as Qwen3.8 27B (Workers AI), the closest-priced model from another provider. - [Claude Fable models](https://codepanel.ai/models/anthropic/claude-fable): Claude Fable has 2 versions CodePanel prices, released from 2026-06-09 to 2026-09-01. The newest, Claude Fable 5.1, costs $10.00 in and $50.00 out per 1M tokens, $0.3829 for one review call (26.2K tokens in, 2.4K out). - [Claude Opus models](https://codepanel.ai/models/anthropic/claude-opus): Claude Opus has 5 versions CodePanel prices, released from 2026-02-05 to 2026-09-22. The newest, Claude Opus 5.5, costs $4.00 in and $20.00 out per 1M tokens, $0.1532 for one review call (26.2K tokens in, 2.4K out). Across the line, Claude Opus 4.6 to Claude Opus 5.5: input $5.00 to $4.00 and output $25.00 to $20.00 per 1M. - [Claude Sonnet models](https://codepanel.ai/models/anthropic/claude-sonnet): Claude Sonnet has 3 versions CodePanel prices, released from 2026-02-17 to 2026-09-28. The newest, Claude Sonnet 5.5, costs $2.00 in and $10.00 out per 1M tokens, $0.0766 for one review call (26.2K tokens in, 2.4K out). Across the line, Claude Sonnet 4.6 to Claude Sonnet 5.5: input $3.00 to $2.00 and output $15.00 to $10.00 per 1M. - [Claude Haiku models](https://codepanel.ai/models/anthropic/claude-haiku): Claude Haiku has 2 versions CodePanel prices, released from 2025-10-15 to 2026-10-07. The newest, Claude Haiku 5.5, costs $0.10 in and $0.50 out per 1M tokens, $0.0038 for one review call (26.2K tokens in, 2.4K out). - [GPT Sol models](https://codepanel.ai/models/openai/gpt-sol): GPT Sol has 3 versions CodePanel prices, released from 2026-07-09 to 2026-09-29. The newest, GPT-6.1 Sol, costs $2.00 in and $10.00 out per 1M tokens, $0.0766 for one review call (26.2K tokens in, 2.4K out). Across the line, GPT-5.6 Sol to GPT-6.1 Sol: input $4.00 to $2.00 and output $20.00 to $10.00 per 1M. - [GPT Luna models](https://codepanel.ai/models/openai/gpt-luna): GPT Luna has 2 versions CodePanel prices, released from 2026-07-09 to 2026-09-22. The newest, GPT-6 Luna, costs $0.10 in and $0.50 out per 1M tokens, $0.0038 for one review call (26.2K tokens in, 2.4K out). - [GPT models](https://codepanel.ai/models/openai/gpt): GPT has 4 versions CodePanel prices. GPT-5.5 costs $5.00 in and $30.00 out per 1M tokens, $0.2035 for one review call (26.2K tokens in, 2.4K out). Across the line, GPT-5 to GPT-5.5: input $1.25 to $5.00 and output $10.00 to $30.00 per 1M. - [Gemini Pro models](https://codepanel.ai/models/google/gemini-pro): Gemini Pro has 2 versions CodePanel prices, released from 2025-03-25 to 2026-02-19. The newest, Gemini 3.1 Pro, costs $2.00 in and $12.00 out per 1M tokens, $0.0814 for one review call (26.2K tokens in, 2.4K out). - [Gemini Flash models](https://codepanel.ai/models/google/gemini-flash): Gemini Flash has 5 versions CodePanel prices. Gemini 3.8 Flash costs $0.75 in and $3.75 out per 1M tokens, $0.0287 for one review call (26.2K tokens in, 2.4K out). Across the line, Gemini 3.7 Flash to Gemini 3.8 Flash: the same input and output rates. - [Gemini Flash-Lite models](https://codepanel.ai/models/google/gemini-flash-lite): Gemini Flash-Lite has 3 versions CodePanel prices, released from 2025-06-17 to 2026-07-21. The newest, Gemini 3.5 Flash-Lite, costs $0.30 in and $2.50 out per 1M tokens, $0.0139 for one review call (26.2K tokens in, 2.4K out). Across the line, Gemini 2.5 Flash-Lite to Gemini 3.5 Flash-Lite: input $0.10 to $0.30 and output $0.40 to $2.50 per 1M. - [Grok models](https://codepanel.ai/models/xai/grok): Grok has 5 versions CodePanel prices, released from 2026-03 to 2026-09-21. The newest, Grok 4.7, costs $2.00 in and $6.00 out per 1M tokens, $0.0669 for one review call (26.2K tokens in, 2.4K out). Across the line, Grok 4.20 to Grok 4.7: input $1.25 to $2.00 and output $2.50 to $6.00 per 1M. - [Kimi models](https://codepanel.ai/models/moonshot/kimi): Kimi has 2 versions CodePanel prices, released from 2026-04-20 to 2026-07-16. The newest, Kimi K3, costs $3.00 in and $15.00 out per 1M tokens, $0.1149 for one review call (26.2K tokens in, 2.4K out). - [Kimi Code models](https://codepanel.ai/models/moonshot/kimi-code): Kimi Code has 2 versions CodePanel prices. Kimi K2.7 Code Highspeed costs $1.90 in and $8.00 out per 1M tokens, $0.0691 for one review call (26.2K tokens in, 2.4K out). - [GLM models](https://codepanel.ai/models/zai/glm): GLM has 4 versions CodePanel prices. GLM-5.3 (Workers AI) costs $1.40 in and $4.40 out per 1M tokens, $0.0473 for one review call (26.2K tokens in, 2.4K out). Across the line, GLM-4.7 to GLM-5.3 (Workers AI): input $0.54 to $1.40 and output $1.98 to $4.40 per 1M. - [Qwen Max models](https://codepanel.ai/models/qwen/qwen-max): Qwen Max has 2 versions CodePanel prices. Qwen3.8 Max (OpenRouter) costs $2.00 in and $6.00 out per 1M tokens, $0.0669 for one review call (26.2K tokens in, 2.4K out). - [MiMo Pro models](https://codepanel.ai/models/xiaomi/mimo-pro): MiMo Pro has 2 versions CodePanel prices, released 2026-09-21. The newest, MiMo-V2.6-Pro (OpenRouter), costs $0.435 in and $0.87 out per 1M tokens, $0.0135 for one review call (26.2K tokens in, 2.4K out). - [Muse Spark models](https://codepanel.ai/models/meta/muse-spark): Muse Spark has 2 versions CodePanel prices. Muse Spark 1.2 costs $1.25 in and $4.25 out per 1M tokens, $0.043 for one review call (26.2K tokens in, 2.4K out). - [Mistral Large models](https://codepanel.ai/models/mistral/mistral-large): Mistral Large has 2 versions CodePanel prices. Mistral Large 4 costs $1.36 in and $4.18 out per 1M tokens, $0.0457 for one review call (26.2K tokens in, 2.4K out). - [Cloudflare Workers AI models](https://codepanel.ai/models/hosts/workers-ai): Cloudflare Workers AI serves 9 models CodePanel offers, from OpenAI, Google, DeepSeek, Moonshot, Z.ai, Qwen and NVIDIA: $0.10 to $1.40 per 1M input tokens and $0.30 to $4.40 per 1M output tokens. One review call (26.2K tokens in, 2.4K out) costs from $0.0033 on Gemma 4 26B A4B (Workers AI) to $0.0473 on GLM-5.3 (Workers AI). - [OpenRouter models](https://codepanel.ai/models/hosts/openrouter): OpenRouter serves 17 models CodePanel offers, from Qwen, StepFun, Xiaomi, Google, Meta, Mistral, OpenAI, Thinking Machines and Z.ai: $0.0341 to $4.35 per 1M input tokens and $0.87 to $14.00 per 1M output tokens. One review call (26.2K tokens in, 2.4K out) costs from $0.0135 on MiMo-V2.6-Pro (OpenRouter) to $0.135 on MiMo-V2.6-Pro UltraSpeed (OpenRouter). - [Economy LLMs: under $0.02 a review call](https://codepanel.ai/models/tiers/economy): 23 models CodePanel offers fall in the economy tier: under $0.02 for one review call. One review call (26.2K tokens in, 2.4K out) costs from $0.0027 on GPT-OSS 20B (Groq) to $0.0195 on Qwen3.8 27B (Workers AI). - [Mid-range LLMs: $0.02 to $0.15 a review call](https://codepanel.ai/models/tiers/mid-range): 43 models CodePanel offers fall in the mid-range tier: $0.02 to $0.15 for one review call. One review call (26.2K tokens in, 2.4K out) costs from $0.0287 on Gemini 3.8 Flash to $0.135 on MiMo-V2.6-Pro UltraSpeed (OpenRouter). - [Frontier LLMs: $0.15 or more a review call](https://codepanel.ai/models/tiers/frontier): 10 models CodePanel offers fall in the frontier tier: $0.15 or more for one review call. One review call (26.2K tokens in, 2.4K out) costs from $0.1532 on Claude Opus 5.5 to $0.3829 on Claude Fable 5.1. - [Open-weight LLMs: benchmark, pricing and licenses](https://codepanel.ai/models/open-weight): 23 open-weight models from OpenAI, Google, DeepSeek, Moonshot, Mistral, Qwen, Z.ai, NVIDIA, StepFun and Xiaomi can review pull requests in CodePanel, called on Cloudflare Workers AI, DeepSeek, Moonshot (Kimi), Mistral, Groq and OpenRouter: $0.075 to $4.35 per 1M input tokens and $0.30 to $15.00 per 1M output tokens. One review call (26.2K tokens in, 2.4K out) costs from $0.0027 on GPT-OSS 20B (Groq) to $0.135 on MiMo-V2.6-Pro UltraSpeed (OpenRouter). The weights are public, so a team can also run them on its own hardware. ## Agents - [Coding agents in the review loop](https://codepanel.ai/agents.md): Every coding agent and assistant CodePanel works with, how each one connects, and the instruction files they read. - [Code review with Claude Code](https://codepanel.ai/agents/claude-code.md): How Claude Code runs the review loop with CodePanel: it waits for the review of its own commit, fixes or disputes each finding, and rates every one. - [Claude Code hooks for code review](https://codepanel.ai/agents/claude-code/hooks.md): A Stop hook that keeps Claude Code working while CodePanel's merge gate has not passed on the pull request's head commit, tested on each path it can take. - [Code review with Codex](https://codepanel.ai/agents/codex.md): Codex, OpenAI's coding agent, in CodePanel's review loop: it connects through MCP and skills, reads AGENTS.md and answers every finding. - [Code review with Cursor](https://codepanel.ai/agents/cursor.md): Cursor, Anysphere's coding agent, in CodePanel's review loop: it connects through MCP and skills, reads Cursor rules and answers every finding. - [Code review with Copilot in VS Code](https://codepanel.ai/agents/vs-code.md): Copilot in VS Code, GitHub's coding agent, in CodePanel's review loop: it connects through MCP and a setup prompt, reads AGENTS.md and answers every finding. - [Code review with Gemini CLI](https://codepanel.ai/agents/gemini-cli.md): Gemini CLI, Google's coding agent, in CodePanel's review loop: it connects through MCP and skills, reads GEMINI.md and answers every finding. - [Code review with OpenCode](https://codepanel.ai/agents/opencode.md): OpenCode, Anomaly's coding agent, in CodePanel's review loop: it connects through MCP and skills, reads AGENTS.md and answers every finding. - [Code review with GitHub Copilot CLI](https://codepanel.ai/agents/copilot-cli.md): GitHub Copilot CLI, GitHub's coding agent, in CodePanel's review loop: it connects through MCP and skills, reads AGENTS.md and answers every finding. - [CodePanel in Claude](https://codepanel.ai/agents/claude-apps.md): Claude reads CodePanel reviews and rates findings in chat through a remote MCP server, while a coding agent pushes the fixes. - [CodePanel in ChatGPT](https://codepanel.ai/agents/chatgpt.md): ChatGPT reads CodePanel reviews and rates findings in chat through a remote MCP server, while a coding agent pushes the fixes. - [AGENTS.md for an agent-to-agent review loop](https://codepanel.ai/agents/files/agents-md.md): What to put in AGENTS.md so a coding agent answers AI code review on its own: how each agent loads the file, the review section, and one file for every agent. - [CLAUDE.md for an agent-to-agent review loop](https://codepanel.ai/agents/files/claude-md.md): What to put in CLAUDE.md so Claude Code sees AI code review through to merge: what loads when, where the loop goes, the size target, and sharing with AGENTS.md. - [copilot-instructions.md for an agent-to-agent review loop](https://codepanel.ai/agents/files/copilot-instructions.md): How GitHub Copilot CLI and Copilot in VS Code combine copilot-instructions.md with AGENTS.md, CLAUDE.md and GEMINI.md, and where the review loop belongs. - [Cursor rules for an agent-to-agent review loop](https://codepanel.ai/agents/files/cursor-rules.md): Cursor rules for AI code review: which .mdc rule type to use, scoping a rule with globs like a .codepanel/rules.md section, and when to use AGENTS.md. - [GEMINI.md for an agent-to-agent review loop](https://codepanel.ai/agents/files/gemini-md.md): What Gemini CLI loads from GEMINI.md, why it skips AGENTS.md by default, and three ways to give it the same review instructions as every other agent. ## Glossary - [Code review glossary](https://codepanel.ai/glossary.md): Terms from code review between coding agents and reviewer agents, each defined in a sentence and shown in a real review loop. - [Blocking finding](https://codepanel.ai/glossary/blocking-finding.md): A review finding at or above the repository's gate level that has no answer yet. While one is open, the merge gate fails. - [False positive](https://codepanel.ai/glossary/false-positive.md): A review finding that names a problem the code does not have. With missed bugs, it is the main cost of an AI code reviewer. - [LGTM](https://codepanel.ai/glossary/lgtm.md): LGTM means "looks good to me": a reviewer's sign-off that a pull request can merge as it is, written in a comment or with an approving review. - [Merge gate](https://codepanel.ai/glossary/merge-gate.md): A required check that keeps a pull request from merging until its review findings are dealt with, so a coding agent knows when it is done. - [Nit](https://codepanel.ai/glossary/nit.md): A nit, short for nitpick, is a minor review comment the author is free to ignore, such as a variable name, a comment's wording or a long line. - [Orchestrator](https://codepanel.ai/glossary/orchestrator.md): In a panel review, the model that merges every reviewer's findings, drops what the code contradicts and writes the one review that is posted. ## Optional - [Privacy Policy](https://codepanel.ai/privacy): What CodePanel stores, what it sends to AI providers, and how to delete it. - [Terms of Service](https://codepanel.ai/terms): The terms for using CodePanel.