Benchmarks / Professional Domains
TaxEval
Vals AI's tax-law reasoning benchmark.
maintained by Vals AI · www.vals.ai · 131 observations on record
Versions
1 version
| Version | Tasks |
|---|---|
| TaxEval current tax-eval | — |
Current leaderboard
TaxEval · Accuracy · top 25 of 131 systems
| # | System | Accuracy % |
|---|---|---|
| 1 | Muse Spark 1.1xhigh effort Meta AI | 79.7% |
| 2 | Muse Spark Meta AI | 77.7% |
| 3 | Claude Sonnet 4.6max effort Anthropic | 77.1% |
| 4 | Claude Fable 5max effort Anthropic | 76.9% |
| 5 | GPT 5.6 Terraxhigh effort OpenAI | 76.2% |
| 6 | GPT 5.6 Lunamax effort OpenAI | 76.2% |
| 7 | Claude Opus 4.6max effort Anthropic | 76.0% |
| 8 | Grok 3 xAI | 75.9% |
| 9 | GPT 5.2 (2025-12-11)xhigh effort OpenAI | 75.8% |
| 10 | Grok 4 Fast Reasoning xAI | 75.7% |
| 11 | Claude Sonnet 5max effort Anthropic | 75.6% |
| 12 | Claude Opus 4.8max effort Anthropic | 75.6% |
| 13 | Qwen3 7 Max Alibaba (Qwen) | 75.3% |
| 14 | Claude Opus 4.7max effort Anthropic | 75.3% |
| 15 | GPT 5 Mini (2025-08-07)high effort OpenAI | 75.2% |
| 16 | GPT 4.1 (2025-04-14)high effort OpenAI | 75.1% |
| 17 | GPT 5.5 (2026-04-22)xhigh effort OpenAI | 75.0% |
| 18 | GPT 5.1 (2025-11-13)high effort OpenAI | 74.9% |
| 19 | Claude Opus 4.5 20251101high effort Anthropic | 74.9% |
| 20 | O4 Mini (2025-04-16)high effort OpenAI | 74.8% |
| 21 | GPT 5.6 Solmax effort OpenAI | 74.8% |
| 22 | Qwen3 6 Plus Alibaba | 74.7% |
| 23 | Kimi K2 6 Moonshot AI | 74.7% |
| 24 | O3 (2025-04-16)high effort OpenAI | 74.6% |
| 25 | GPT 4o (2024-11-20)high effort OpenAI | 74.5% |
One row per evaluated system — reasoning-effort variants rank separately.