Tax Agent Bench — Rule & Source Lookup
Professional Work · 2026-09-23
Weighted rubric and citation credit on 30 rule-and-source tax questions, using the parent evaluation protocol.
Top models (higher is better)
| Model | Score |
|---|---|
| Qwen3.8 Max (0902) | 81.5 |
| GLM-5.3 | 81.3 |
| Claude Opus 5 | 78.0 |
| Gemini 4 Argon | 77.4 |
| Muse Spark 1.3 | 74.7 |
| Gemini 4 Carbon | 74.0 |
| Claude Fable 5.1 | 72.9 |
| DeepSeek V4.1 Flash 1017 | 70.8 |
| DeepSeek-V4-Flash-0731 | 70.2 |
| Muse Wonder Full | 68.1 |
| Qwen3.8 27B | 68.0 |
| Muse Wonder | 67.7 |
| Grok 4.6 | 67.5 |
| Claude Haiku 5.5 | 66.4 |
| Sonnet 4.6 | 65.9 |
| Sonnet 5 | 61.5 |
| Opus 4.7 | 61.0 |
| Mistral Large 4 Preview | 60.8 |
| Step 5 Preview | 60.8 |
| Grok 4.5 | 60.6 |
| Muse Spark Open | 60.3 |
| Gemini 3.8 Flash | 60.2 |
| GLM-5.3 Flash | 60.2 |
| MiMo-V2.6-Pro | 59.8 |
| Kimi K3 | 58.4 |
| Gemini 3.7 Flash | 57.9 |
| MiniMax M3 | 57.9 |
| MiniMax M3.1 Flash | 57.7 |
| Hy4 Preview | 57.5 |
| DeepSeek V4.1 Flash | 56.9 |
| MiMo-V2.6-Flash | 56.3 |
| Claude Fable 5 | 55.9 |
| Muse Spark 1.2 | 55.1 |
| Opus 4.8 | 54.7 |
| GPT-5.6 Luna | 54.3 |
| GPT-5.6 Sol | 53.8 |
| Kimi K3.1 Pro | 53.8 |
| GPT-5.6 Terra | 53.3 |
| Grok 4.7 | 52.2 |
| Ember-1 | 51.2 |