AutomationBench-AA Finance
Agents
AutomationBench-AA on its finance-domain workflows across business apps such as spreadsheets, email and accounting tools, as published in the Artificial Analysis Finance & Accounting Index.
Top models (higher is better)
| Model | Score |
|---|---|
| Gemini 4 Carbon | 86.1 |
| Kimi K3.1 Pro | 80.3 |
| MiniMax M3.1 Flash | 79.5 |
| Claude Sonnet 5.5 | 79.4 |
| Ling 3.1 Flash 1019 | 78.4 |
| DeepSeek V4.1 Flash 1017 | 74.1 |
| Grok 4.7 | 72.7 |
| GPT-6 Astra | 69.5 |
| Muse Wonder Full | 65.5 |
| Muse Wonder | 63.1 |
| GLM-5.3 | 62.3 |
| Gemini 3.8 Flash | 59.4 |
| GPT-6.1 Sol | 59.2 |
| Qwen3.8 Max (0902) | 56.6 |
| MiMo-V2.6-Pro | 56.3 |
| Mistral Large 4 | 54.5 |
| GPT-6.1 Terra | 54.4 |
| Muse Spark Open | 44.2 |
| Claude Haiku 5.5 | 43.2 |
| Step 5 Preview | 42.8 |
| Beam 1.0 501B A23B | 42.4 |
| GPT-6 Luna | 41.0 |
| Qwen3.8 27B | 41.0 |
| K2 Horizon 375B A23B | 30.4 |
| Nemotron 3.5 Super VL 120B A12B (precision unspecified) | 10.3 |
| Inkling | 5.1 |