TaxBench Tax Calculations pass^5
Professional Work · 2026-05-04
Tax Calculations subset of TaxBench reporting pass^5 repeated-attempt reliability performance on professional tax tasks.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-5.4 Pro | 27.5 |
| GPT-5.5 Pro | 22.6 |
| GPT-5.5 | 21.1 |
| Opus 4.7 | 19.7 |
| Gemini 3.1 Pro Preview | 18.4 |
| Opus 4.6 | 13.8 |
| Grok 4.1 Fast | 12.9 |
| Gemini 3.1 Flash (benchmark label) | 12.0 |
| Sonnet 4.6 | 8.7 |
| GPT-5.2 Pro | 8.0 |
| Sonnet 4.5 | 6.3 |
| Gemini 2.5 Pro | 5.7 |
| GPT-5.4 | 5.2 |
| Grok 4.20 | 5.2 |
| GPT-5.2 | 1.6 |