AutomationBench-AA: leaderboard
Artificial Analysis evaluates agents carrying out workplace objectives subject to guardrails. The headline score adjusts objective completion for violations; the strict board requires full task success. Both are percentages from the complete public model roster.
Metric: Guardrail-Adjusted Objective Completion (%). Source: artificialanalysis.ai. Saturation forecast: Around February 2027. 243 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 4 Carbon (High) | 77.82 | #1 (Gemini 4 Carbon) |
| 2 | Gemini 4 Argon (High) | 77.51 | #9 (Gemini 4 Argon) |
| 3 | Kimi K3.1 Pro | 73.3 | #17 |
| 4 | MiniMax-M3.1-Flash | 72.52 | #52 |
| 5 | Muse Wonder Full (Max) | 71.33 | #12 (Muse Wonder Full) |
| 6 | DeepSeek V4.1 Flash (Max) | 68.89 | #37 (DeepSeek V4.1 Flash) |
| 7 | GPT-6 Astra (Max) | 68.49 | #6 (GPT-6 Astra) |
| 8 | DeepSeek V4.1 Flash (1017) (High) | 68.49 | #27 (DeepSeek V4.1 Flash (1017)) |
| 9 | Muse Wonder (Max) | 68.09 | #18 (Muse Wonder) |
| 10 | GPT-6 Astra (xHigh) | 67.18 | #6 (GPT-6 Astra) |
| 11 | Grok 4.6 (xHigh) | 67 | #29 (Grok 4.6) |
| 12 | Grok 4.6 (High) | 66.68 | #29 (Grok 4.6) |
| 13 | DeepSeek V4.1 Flash (1017) (Max) | 66.64 | #27 (DeepSeek V4.1 Flash (1017)) |
| 14 | GPT-6 Astra (High) | 66.61 | #6 (GPT-6 Astra) |
| 15 | GPT-6.1 Sol (xHigh) | 66.57 | #7 (GPT-6.1 Sol) |
No result here: #3 Claude Opus 5.5, #10 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=automationbench-aa · How It Works · Data refreshed daily, snapshot 2026-11-05.