Go code review: which models catch the bugs
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%.
| Model | Caught here | Every bug | Difference |
|---|---|---|---|
| GPT-6.1 Sol | 75%[25%–100%] | 89% | −14 pts |
| Gemini 3.8 Flash | 75%[25%–100%] | 68% | +7 pts |
| Claude Haiku 5.5 | 75%[25%–100%] | 64% | +11 pts |
| Claude Opus 5.5 | 50%[0%–100%] | 90% | −40 pts |
| Gemini 4 Argon | 50%[0%–100%] | 82% | −32 pts |
| Claude Sonnet 5.5 | 50%[0%–100%] | 77% | −27 pts |
| GPT-5.6 Sol | 50%[0%–100%] | 72% | −22 pts |
| Gemini 3.1 Pro | 38%[0%–75%] | 67% | −29 pts |
| GLM-5.3 Flash (Workers AI) | 31%[0%–63%] | 61% | −30 pts |
| Kimi K2.7 Code (Workers AI) | 26%[7%–44%] | 38% | −12 pts |
| GPT-6 Luna | 25%[0%–75%] | 67% | −42 pts |
| DeepSeek V4 Pro (Workers AI) | 25%[0%–75%] | 54% | −29 pts |
| DeepSeek V4 Flash (Workers AI) | 25%[0%–56%] | 49% | −24 pts |
| GLM-5.3 (Workers AI) | 19%[0%–38%] | 61% | −42 pts |
| GPT-OSS 120B (Workers AI) | 13%[0%–25%] | 40% | −27 pts |
| Gemma 4 26B A4B (Workers AI) | 0% | 26% | −26 pts |
| Gemini 3.5 Flash-Lite | 0% | 11% | −11 pts |
On average these models caught 37% of known bugs in Go code, less often than the 60% the same 17 caught across the whole benchmark: the hardest of the 12 languages.
Claude Haiku 5.5 gains the most here: 75%, against 64% over every bug.
GPT-6 Luna drops the most: 25%, against 67% over every bug.
The change “feat: middleware.ClientIP, a replacement for middleware.RealIP” in go-chi/chi, a change of over 300 lines that landed in May 2026, introduced a security bug that a later upstream commit fixed.
GPT-6.1 Sol, Gemini 4 Argon, Claude Sonnet 5.5, GPT-5.6 Sol, Gemini 3.8 Flash and GPT-6 Luna caught it; Claude Opus 5.5, Claude Haiku 5.5, Gemini 3.1 Pro, GLM-5.3 Flash, GLM-5.3, DeepSeek V4 Pro (Workers AI), DeepSeek V4 Flash (Workers AI), GPT-OSS 120B (Workers AI), Kimi K2.7 Code (Workers AI), Gemma 4 26B A4B and Gemini 3.5 Flash-Lite missed it.