NL2Repo-Bench
Can the model complete substantial programming work under this benchmark's agent setup?
Coding9 ranked models9 reported values2 reportsHigher is better
Top rankings
One row per model, using its best reported score across effort settings.
| Rank | Model | Best score | Best reported setting |
|---|---|---|---|
| 1 | Claude Opus 4.8Anthropic | 69.7% | Reported configurationZ.ai report1 value · 1 reportJun 16, 2026 · source |
| 2 | DeepSeek V4.1 FlashDeepSeek | 65.4% | Reported configurationDeepSeek report1 value · 1 reportSep 10, 2026 · source |
| 3 | GPT-5.5OpenAI | 50.7% | Reported configurationZ.ai report1 value · 1 reportJun 16, 2026 · source |
| 4 | GLM-5.2Z.ai | 48.9% | max effortZ.ai report1 value · 1 reportJun 16, 2026 · source |
| 5 | Qwen3.7-MaxQwen | 47.2% | Reported configurationZ.ai report1 value · 1 reportJun 16, 2026 · source |
| 6 | GLM-5.1Z.ai | 42.7% | Reported configurationZ.ai report1 value · 1 reportJun 16, 2026 · source |
| 7 | MiniMax M3MiniMax | 42.1% | Reported configurationZ.ai report1 value · 1 reportJun 16, 2026 · source |
| 8 | DeepSeek V4 ProDeepSeek | 35.5% | Reported configurationZ.ai report1 value · 1 reportJun 16, 2026 · source |
| 9 | Gemini 3.1 ProGoogle | 33.4% | Reported configurationZ.ai report1 value · 1 reportJun 16, 2026 · source |
Effort curve
Every sourced cost-linked effort value for this exact version. Lines connect complete sweeps only.
No cost-linked effort sweep for this version.
Definition and comparison boundarypublic methodology
Can the model complete substantial programming work under this benchmark's agent setup?
Software implementation, debugging, or repository work under the published evaluation protocol. Higher is better. The value is the percentage reported in this lab's table.
Harness, tools, prompts, attempt count, and benchmark version can materially change the result. Compare within one reporting context.
DeepSeek's official V4.1-Flash release table; exact source benchmark label NL2Repo-Bench retained.
Method / source