TestsBenchmarks · Knowledge
LegalBench (Vals)
Open-source legal reasoning tasks (issue spotting, rules, interpretation, rhetoric).
% solvedHigher is betterToo easy now: the top models all score near perfectSaturatedThe test's websiteOfficial page ↗
Measured by independent testersIndependent Reported by the makerVendor-reported
| Thinking levelSetting | |||||||
|---|---|---|---|---|---|---|---|
| 1 | Claude Fable 5 | Anthropic | 88.6% | Maxeffort=max | Independent testIndependentVals.ai ↗ | — | 29 Sep 2026 |
| 2 | Claude Fable 5.1 | Anthropic | 88.5% | Maxeffort=max | Independent testIndependentVals.ai ↗ | — | 29 Sep 2026 |
| 3 | Gemini 4 Argon | Google (Gemini / DeepMind) | 88.3% | Highreasoning_effort=high | Independent testIndependentVals.ai ↗ | — | 29 Sep 2026 |
| 4 | Gemini 3.1 Pro (Preview) | Google (Gemini / DeepMind) | 87.4% | Highreasoning_effort=high | Independent testIndependentVals.ai ↗ | — | 29 Sep 2026 |
| 5 | Gemini 3.7 Flash | Google (Gemini / DeepMind) | 87.3% | Highreasoning_effort=high | Independent testIndependentVals.ai ↗ | — | 29 Sep 2026 |
| 6 | Gemini 3 Pro | Google (Gemini / DeepMind) | 87.0% | Highreasoning_effort=high | Independent testIndependentVals.ai ↗ | — | 29 Sep 2026 |
| 7 | Gemini 3.8 Flash | Google (Gemini / DeepMind) | 87.0% | Highreasoning_effort=high | Independent testIndependentVals.ai ↗ | — | 29 Sep 2026 |
| 8 | Claude Opus 5 | Anthropic | 87.0% | Maxeffort=max | Independent testIndependentVals.ai ↗ | — | 29 Sep 2026 |
| 9 | GPT-5.6 Sol | OpenAI | 87.0% | Maxreasoning_effort=max | Independent testIndependentVals.ai ↗ | — | 29 Sep 2026 |
| 10 | Gemini 3 Flash Preview | Google (Gemini / DeepMind) | 86.9% | Highreasoning_effort=high | Independent testIndependentVals.ai ↗ | — | 29 Sep 2026 |
| 11 | Gemini 3.6 Flash | Google (Gemini / DeepMind) | 86.7% | Highreasoning_effort=high | Independent testIndependentVals.ai ↗ | — | 29 Sep 2026 |
| 12 | GPT-5.5 | OpenAI | 86.5% | Extra highreasoning_effort=xhigh | Independent testIndependentVals.ai ↗ | — | 29 Sep 2026 |
| 13 | Grok 4.6 | SpaceXAI (formerly xAI) | 86.3% | Highreasoning_effort=high | Independent testIndependentVals.ai ↗ | — | 29 Sep 2026 |
| 14 | GPT-5.4 | OpenAI | 86.0% | Extra highreasoning_effort=xhigh | Independent testIndependentVals.ai ↗ | — | 29 Sep 2026 |
| 15 | Kimi K3 | Moonshot AI (Kimi) | 86.0% | Defaultvals id kimi/kimi-k3 | Independent testIndependentVals.ai ↗ | — | 29 Sep 2026 |
| 16 | Grok 4.5 | SpaceXAI (formerly xAI) | 86.0% | Highreasoning_effort=high | Independent testIndependentVals.ai ↗ | — | 29 Sep 2026 |
| 17 | GPT-5.1 | OpenAI | 85.7% | Highreasoning_effort=high | Independent testIndependentVals.ai ↗ | — | 29 Sep 2026 |
| 18 | MiniMax M3 | MiniMax | 85.4% | Defaultvals id minimax/MiniMax-M3 | Independent testIndependentVals.ai ↗ | — | 29 Sep 2026 |
| 19 | Claude Opus 4.6 | Anthropic | 85.3% | Maxeffort=max | Independent testIndependentVals.ai ↗ | — | 29 Sep 2026 |
| 20 | GLM-5.3 | Z.ai (Zhipu) | 84.8% | Maxreasoning_effort=max | Independent testIndependentVals.ai ↗ | — | 29 Sep 2026 |
| 21 | Grok 4.7 | SpaceXAI (formerly xAI) | 84.4% | Extra highreasoning_effort=xhigh | Independent testIndependentVals.ai ↗ | — | 29 Sep 2026 |
| 22 | GLM-5.3-Flash | Z.ai (Zhipu) | 83.9% | Maxreasoning_effort=max | Independent testIndependentVals.ai ↗ | — | 29 Sep 2026 |
| 23 | Claude Sonnet 5 | Anthropic | 83.9% | Maxeffort=max | Independent testIndependentVals.ai ↗ | — | 29 Sep 2026 |
| 24 | Qwen3.8 Max (0803) | Qwen (Alibaba) | 83.6% | Defaultvals id alibaba/qwen3.8-max | Independent testIndependentVals.ai ↗ | — | 29 Sep 2026 |
| 25 | DeepSeek V4.1 Flash | DeepSeek | 83.3% | Highreasoning_effort=high | Independent testIndependentVals.ai ↗ | — | 29 Sep 2026 |
| 26 | Qwen3.8 27B | Qwen (Alibaba) | 82.4% | Extra highreasoning_effort=xhigh | Independent testIndependentVals.ai ↗ | — | 29 Sep 2026 |
| 27 | DeepSeek V4 Pro (0813) | DeepSeek | 82.4% | Maxreasoning_effort=max | Independent testIndependentVals.ai ↗ | — | 29 Sep 2026 |
| 28 | DeepSeek V4 Pro (Preview, 0423) | DeepSeek | 80.3% | Maxreasoning_effort=max | Independent testIndependentVals.ai ↗ | — | 29 Sep 2026 |
| 29 | DeepSeek V4 Flash (0731) | DeepSeek | 77.7% | Highreasoning_effort=high | Independent testIndependentVals.ai ↗ | — | 29 Sep 2026 |