Automation-Bench
Can the model plan, use tools, and finish a multi-step task?
Agents11 ranked models17 reported values4 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 | DeepSeek V4.1 FlashDeepSeek | 54.8% | Reported configurationDeepSeek report1 value · 1 reportSep 10, 2026 · source |
| 2 | Kimi K3Moonshot AI | 30.8% | max effortMoonshot AI report1 value · 1 reportJul 17, 2026 · source |
| 3 | GPT-5.6 SolOpenAI | 29.7% | max effortMoonshot AI report2 values · 2 reportsJul 17, 2026 · source |
| 4 | Claude Fable 5Anthropic | 29.1% | max effortMoonshot AI report2 values · 2 reportsJul 17, 2026 · source |
| 5 | Claude Opus 4.8Anthropic | 27.2% | max effortMoonshot AI report3 values · 3 reportsJul 17, 2026 · source |
| 6 | GPT-5.5OpenAI | 22.7% | xhigh effortMoonshot AI report3 values · 3 reportsJul 17, 2026 · source |
| 7 | GPT-5.6 TerraOpenAI | 15.2% | Reported configurationOpenAI report1 value · 1 reportJul 9, 2026 · source |
| 8 | GPT-5.6 LunaOpenAI | 14.9% | Reported configurationOpenAI report1 value · 1 reportJul 9, 2026 · source |
| 9 | Gemini 3.5 FlashGoogle | 14.5% | Reported configurationOpenAI report1 value · 1 reportJul 9, 2026 · source |
| 10 | GLM-5.2Z.ai | 12.9% | max effortMoonshot AI report1 value · 1 reportJul 17, 2026 · source |
| 11 | Gemini 3.1 Pro PreviewGoogle | 9.6% | Reported configurationAnthropic report1 value · 1 reportJun 9, 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 plan, use tools, and finish a multi-step task?
Agentic execution across tools, environments, or long-running workflows. 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 label Automation-Bench retained; benchmark-maintainer contexts remain separate.
Method / source