Tax Agent Bench — Current & Temporal Analysis
Professional Work · 2026-09-23
Weighted rubric and citation credit on 30 current-and-temporal tax questions, using the parent evaluation protocol.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Fable 5.1 | 81.2 |
| GLM-5.3 | 78.2 |
| Grok 4.7 | 74.5 |
| GPT-5.6 Sol | 74.4 |
| MiMo-V2.6-Pro | 73.9 |
| Grok 4.6 | 73.4 |
| Claude Opus 5.5 | 73.3 |
| Muse Spark 1.3 | 73.1 |
| Kimi K3 | 72.9 |
| DeepSeek V4.1 Flash | 71.9 |
| Claude Opus 5 | 71.6 |
| Gemini 3.7 Flash | 68.8 |
| GPT-5.5 | 68.2 |
| DeepSeek V4 Pro 0813 | 67.9 |
| GPT-6 Astra | 66.2 |
| Gemini 3.8 Flash | 64.8 |
| Hy4 Preview | 63.7 |
| MiMo-V2.6-Flash | 61.4 |
| GPT-5.6 Terra | 60.3 |
| GPT-5.6 Luna | 60.1 |
| Sonnet 5 | 60.0 |
| GPT-6 Sol | 56.1 |
| GPT-6 Luna | 55.3 |
| MiniMax M3 | 52.9 |
| Inkling | 34.4 |
| Mercury 2.5 | 13.2 |