TaxBench Tax Calculations pass@1
Professional Work · 2026-05-04
Tax Calculations subset of TaxBench reporting pass@1 first-attempt performance on professional tax tasks.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-5.4 Pro | 77.2 |
| GPT-5.5 Pro | 74.3 |
| GPT-5.5 | 73.3 |
| Opus 4.7 | 72.3 |
| Gemini 3.1 Pro Preview | 71.3 |
| Opus 4.6 | 67.3 |
| Grok 4.1 Fast | 66.3 |
| Gemini 3.1 Flash (benchmark label) | 65.4 |
| Sonnet 4.6 | 61.4 |
| GPT-5.2 Pro | 60.4 |
| Sonnet 4.5 | 57.4 |
| Gemini 2.5 Pro | 56.4 |
| GPT-5.4 | 55.5 |
| Grok 4.20 | 55.4 |
| GPT-5.2 | 43.6 |