TaxBench Total pass^5
Professional Work · 2026-05-04
Overall TaxBench pass^5 reliability score, computed across tax knowledge, tax calculations, and data retrieval task groups. The metric emphasizes consistent success across repeated attempts.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-5.5 Pro | 29.3 |
| GPT-5.4 Pro | 27.0 |
| GPT-5.5 | 24.4 |
| Opus 4.6 | 21.4 |
| Gemini 3.1 Pro Preview | 20.1 |
| Grok 4.1 Fast | 19.6 |
| GPT-5.2 Pro | 17.8 |
| Gemini 3.1 Flash (benchmark label) | 15.9 |
| Grok 4.20 | 15.2 |
| Opus 4.7 | 14.4 |
| Sonnet 4.6 | 11.2 |
| GPT-5.4 | 9.3 |
| Gemini 2.5 Pro | 9.0 |
| Sonnet 4.5 | 8.0 |
| GPT-5.2 | 4.6 |