TaxBench Tax Knowledge pass^5
Professional Work · 2026-05-04
Tax Knowledge subset of TaxBench reporting pass^5 repeated-attempt reliability performance on professional tax tasks.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-5.5 Pro | 42.3 |
| Grok 4.1 Fast | 42.3 |
| GPT-5.4 Pro | 40.1 |
| GPT-5.2 Pro | 38.2 |
| Grok 4.20 | 36.4 |
| Gemini 3.1 Flash (benchmark label) | 34.4 |
| GPT-5.5 | 31.3 |
| Opus 4.6 | 31.1 |
| Gemini 3.1 Pro Preview | 29.4 |
| Gemini 2.5 Pro | 20.0 |
| Sonnet 4.6 | 17.8 |
| GPT-5.4 | 16.8 |
| Opus 4.7 | 12.3 |
| Sonnet 4.5 | 11.0 |
| GPT-5.2 | 9.5 |