AA-AnalystAgent (pass^5)
Professional Work
Artificial Analysis' private 80-question spreadsheet and document analysis benchmark across 14 domains. The headline pass^5 is the fraction of questions answered correctly on all five independent attempts with the Stirrup agent harness.
Top models (higher is better)
| Model | Score |
|---|---|
| Gemini 3.7 Flash | 60.0 |
| Claude Fable 5.1 | 57.5 |
| Claude Opus 5 | 53.8 |
| GPT-6 Astra | 51.3 |
| GPT-5.5 | 50.0 |
| Claude Fable 5 | 48.8 |
| GPT-5.6 Sol | 47.5 |
| Sonnet 5 | 46.3 |
| Opus 4.8 | 45.0 |
| Gemini 3.5 Flash | 45.0 |
| Opus 4.7 | 43.8 |
| Gemini 3.1 Pro Preview | 41.3 |
| Grok 4.6 | 41.3 |
| Kimi K3 | 38.8 |
| Grok 4.5 | 35.0 |
| Inkling-Small | 27.5 |
| DeepSeek-V4-Flash | 25.0 |
| Inkling | 23.8 |
| Sonnet 4.6 | 20.0 |
| MiMo-V2.5-Pro | 20.0 |
| DeepSeek-V4-Pro | 18.8 |
| Qwen3.7-Max | 18.8 |
| Ling 3.0 Flash Fin | 16.3 |
| Haiku 4.5 | 15.0 |
| GPT-5.4 Mini | 15.0 |
| Mistral Medium 3.5 | 12.5 |
| MiniMax M2.7 | 11.3 |
| MiniMax M3 | 10.0 |
| Gemini 3.1 Flash-Lite Preview | 8.8 |
| Grok 4.3 | 8.8 |
| Nemotron 3 Ultra 550B A55B | 6.3 |
| Mistral Small 4 | 1.3 |