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).
Side by side
| Figure | Gemini 3.5 Flash-Lite | GLM-5.3 Flash (Workers AI) |
|---|---|---|
| Price tier | Economy | Economy |
| Bugs caughtshare of known bugs | 11% | 61% (better) |
| False alarmsper clean PR | 0.4 (better) | 1.8 |
| Cost per reviewmeasured | $0.0059 | $0.0035 (better) |
| Cost per bug caught | $0.0528 | $0.0066 (better) |
| Exact linecatches on the expected lines | 83% | 89% (better) |
| Review timemean | 103s (better) | 254s |
| Typical review calllist price | $0.0139 | $0.0051 (better) |
| Input / 1M tokens | $0.30 | $0.15 (better) |
| Output / 1M tokens | $2.50 | $0.50 (better) |
| Context window | 1M | 1M |
Head to head
The known bugs both were scored on, by who caught each in at least one run.
By language
| Model | All | TypeScript6 bugs | Go4 bugs | Java4 bugs | Rust4 bugs | C#3 bugs | JavaScript3 bugs | Kotlin3 bugs | PHP3 bugs | Python3 bugs | Ruby3 bugs | Swift3 bugs | C++2 bugs | C1 bug |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Gemini 3.5 Flash-Lite | 11% | 33% | 0% | 5% | 0% | 0% | 0% | 33% | 0% | 17% | 0% | 0% | 50% | 0% |
| GLM-5.3 Flash (Workers AI) | 61% | 57% | 31% | 45% | 88% | 30% | 67% | 89% | 33% | 78% | 67% | 67% | 100% | 67% |
Of the 42 known bugs both were scored on, both caught 6 and neither caught 7. GLM-5.3 Flash (Workers AI) caught 29 bugs that Gemini 3.5 Flash-Lite missed: 4 in Rust, 3 in Java, 3 in JavaScript, 3 in Ruby, 3 in Swift, 2 in C#, 2 in Go, 2 in Kotlin, 2 in Python, 2 in TypeScript, 1 in C, 1 in C++ and 1 in PHP.
On pull requests with no known bug, Gemini 3.5 Flash-Lite raised 0.4 false alarms per review and GLM-5.3 Flash (Workers AI) 1.8. A measured review cost $0.0059 on Gemini 3.5 Flash-Lite, 66% more than $0.0035 on GLM-5.3 Flash (Workers AI).
| Pull request | Language · kind | Gemini 3.5 Flash-Lite | GLM-5.3 Flash (Workers AI) |
|---|---|---|---|
| MudBlazor/MudBlazorRefactor: MudDebouncedInput, ThrottleDispatcher and DebounceDispatcher | C# · Concurrency | 0/3 | 2/9 |
| OrchardCMS/OrchardCoreSimplify using `StringValues` in Liquid templates | C# · Security | 0/1 | 2/3 |
| aio-libs/aiohttpkeep websocket compression state across interleaved control frames | Python · Logic | 0/1 | 2/4 |
| better-auth/better-authfix(saml): IdP-Initiated Callback Routing | TypeScript · Security | 0/2 | 4/4 |
| c-ares/c-aresEvent Thread could stall when not notified of new queries on existing connections that are in a bad state | C · Resource handling | 0/1 | 2/3 |
| denoland/denoperf(libs/core): poll only ready handles and yield between I/O batches | Rust · Concurrency | 0/1 | 3/3 |
| django/djangoFixed #36226 -- Accepted str or bytes for password and salt in password hashers. | Python · Error handling | 0/1 | 3/3 |
| ggml-org/llama.cpprpc : cache and reuse compute graphs | C++ · Security | 0/1 | 3/3 |
| golang/gocrypto/tls: add LocalCertificate to ConnectionState | Go · Logic | 0/1 | 3/4 |
| groue/GRDB.swiftDatabaseQueue and DatabaseSnapshot reads can access task locals | Swift · Concurrency | 0/1 | 1/3 |
| iBotPeaches/Apktoolrefactor: Improve renaming/injection of resources in stripped APKs | Java · Security | 0/1 | 2/3 |
| jhy/jsoupStore cached child els in Attributes, not field | Java · Concurrency | 0/2 | 2/6 |
| kotest/kotestFix DataClassEq comparator | Kotlin · Logic | 0/1 | 2/3 |
| ktorio/ktorKTOR-9171 Redesign ByteReadChannel.readUTF8Line API | Kotlin · Logic | 0/1 | 4/4 |
| laravel/framework[12.x] Colocate Container build functions with the `Buildable` interface | PHP · Error handling | 0/1 | 3/3 |
| mockito/mockitoAdd core API to enable Kotlin singleton mocking | Java · API misuse | 0/1 | 1/4 |
| nodejs/undicifeat: add configurable maxPayloadSize for WebSocket | JavaScript · Security | 0/1 | 3/3 |
| nushell/nushellreorganize nushell configuration | Rust · Error handling | 0/1 | 2/4 |
| pinojs/thread-streamrefactor: replace setTimeout with Atomics.waitAsync | JavaScript · Concurrency | 0/1 | 1/3 |
| pointfreeco/swift-composable-architectureRemove child store from parent when invalidated | Swift · Resource handling | 0/1 | 2/3 |
| pointfreeco/swift-navigationUIKit: Fix switch presented child double dismiss | Swift · Logic | 0/1 | 3/3 |
| rack/rackApply bytesize and number of param limits in QueryParser | Ruby · Security | 0/1 | 3/3 |
| rails/railsReimplement Action Cable redis adapter with redis-client | Ruby · Logic | 0/1 | 2/3 |
| rustls/rustlsAffix `Protocol` at top of client handshake | Rust · Logic | 0/1 | 3/3 |
| sindresorhus/p-throttleAdd `weight` option | JavaScript · Logic | 0/1 | 2/3 |
| socketry/asyncBarrier wait in completion order + `Queue#close`. | Ruby · Concurrency | 0/1 | 1/3 |
| tokio-rs/tokiotime: move lazy-registration state into `Sleep` | Rust · Logic | 0/1 | 4/4 |
| traefik/traefikAdd the option to define custom cipher suites for backend serversTransport | Go · API misuse | 0/1 | 2/4 |
| trpc/trpcfix(server): pass maxDurationMs timeout signal to subscription handlers | TypeScript · Concurrency | 0/1 | 2/3 |
Defects caught over all runs: a PR reviewed three times with one known bug shows 3/3 when it was caught every time.
Every bug each was scored on: Gemini 3.5 Flash-Lite and GLM-5.3 Flash (Workers AI).