Tax Agent Bench — Fact-Pattern Analysis
Professional Work · 2026-09-23
Weighted rubric and citation credit on 43 fact-pattern tax questions, using the parent evaluation protocol.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Fable 5.1 | 79.3 |
| Claude Opus 5 | 77.3 |
| Gemini 4 Argon | 76.7 |
| Muse Spark 1.3 | 75.1 |
| Muse Wonder Full | 74.8 |
| Claude Fable 5 | 74.3 |
| DeepSeek V4.1 Flash 1017 | 72.5 |
| Ember-1 | 71.8 |
| Gemini 4 Carbon | 70.4 |
| Qwen3.8 Max (0902) | 70.3 |
| Kimi K3.1 Pro | 68.8 |
| Gemini 3.8 Flash | 68.0 |
| Muse Wonder | 67.6 |
| GPT-6 Astra | 66.0 |
| MiMo-V2.6-Flash | 66.0 |
| Grok 4.6 | 66.0 |
| Opus 4.8 | 65.4 |
| Grok 4.7 | 64.3 |
| GPT-5.6 Terra | 63.3 |
| Kimi K3 | 62.2 |
| GPT-6.1 Sol | 61.7 |
| Opus 4.7 | 61.6 |
| DeepSeek V4.1 Flash | 60.6 |
| GPT-5.6 Sol | 60.4 |
| GPT-6 Luna | 60.2 |
| Sonnet 5 | 59.9 |
| DeepSeek V4 Pro 0813 | 59.7 |
| GPT-5.6 Luna | 59.6 |
| Grok 4.5 | 59.5 |
| Claude Haiku 5.5 | 59.2 |
| GPT-5.5 | 59.2 |
| MiMo-V2.6-Pro | 58.4 |
| GLM-5.3 | 58.2 |
| Gemini 3.5 Flash | 58.1 |
| Qwen3.8 27B | 57.7 |
| GLM-5.3 Flash | 57.3 |
| Mistral Large 4 Preview | 57.2 |
| Sonnet 4.6 | 56.7 |
| Step 5 Preview | 56.4 |
| Hy4 Preview | 55.8 |