ProgramBench
Can the model complete substantial programming work under this benchmark's agent setup?
Top rankings
One row per model, using its best reported score across effort settings.
| Rank | Model | Best score | Best reported setting |
|---|---|---|---|
| 1 | Kimi K3Moonshot AI | 77.8% | max effortMoonshot AI report1 value · 1 reportJul 17, 2026 · source |
| 2 | GPT-5.6 SolOpenAI | 77.6% | max effortMoonshot AI report1 value · 1 reportJul 17, 2026 · source |
| 3 | Claude Fable 5Anthropic | 76.8% | max effortMoonshot AI report1 value · 1 reportJul 17, 2026 · source |
| 4 | Claude Opus 4.8Anthropic | 71.9% | max effortMoonshot AI report3 values · 3 reportsJul 17, 2026 · source |
| 5 | GPT-5.5OpenAI | 70.8% | xhigh effortMoonshot AI report3 values · 3 reportsJul 17, 2026 · source |
| 6 | GLM-5.2Z.ai | 63.7% | max effortMoonshot AI report2 values · 2 reportsJul 17, 2026 · source |
| 7 | Kimi K2.7 CodeMoonshot AI | 53.6% | thinking effortMoonshot AI report1 value · 1 reportJul 22, 2026 · source |
| 8 | GLM-5.1Z.ai | 50.9% | Reported configurationZ.ai report1 value · 1 reportJun 16, 2026 · source |
| 9 | Kimi K2.6Moonshot AI | 48.3% | thinking effortMoonshot AI report1 value · 1 reportJul 22, 2026 · source |
| 10 | DeepSeek V4 ProDeepSeek | 47.8% | Reported configurationZ.ai report1 value · 1 reportJun 16, 2026 · source |
| 11 | Gemini 3.1 ProGoogle | 39.5% | Reported configurationZ.ai report1 value · 1 reportJun 16, 2026 · source |
| 12 | DeepSeek V4.1 FlashDeepSeek | 20.3% | Reported configurationDeepSeek report1 value · 1 reportSep 10, 2026 · source |
Effort curve
Every sourced cost-linked effort value for this exact version. Lines connect complete sweeps only.
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 ProgramBench retained. Kimi Code harness. Claude Code 2.1.156; max reasoning effort for GLM-5.2. Kimi Code CLI with thinking enabled, temperature 1.0, top-p 0.95, and 262,144-token context; GPT-5.5 used Codex xhigh and Claude Opus 4.8 used Claude Code xhigh.
Method / source