Benchmarks · Vals AI
Vals Legal Index
Composite index of legal research and drafting by practice area (business, civil, employment, family, criminal, tax, Legal Research Bench, HLAB), weighted by each area's share of US law-office revenue.
- Models measured
- 36
- Top
- 77.86 %Wonder Fullwithin the standard error of ±1.82
- Version
- 1
- Measured
- 9 November 2026
01
Ranking
At the top: Top 15
Vals AI Legal Research Bench, all-pass [1]. Hatched bars are rejected by the compliance gate.
- 1Wonder Full77.86%
- 2Gemini 4 Argon76.78%
- 3Gemini 4 Carbon75.93%
- 4Muse Ember75.69%
- 5Wonder75.26%
- 6GLM 5.374.62%
- 7Inkling 2 Small73.89%
- 8Grok 4.772.67%
- 9Claude Opus 5.572.37%
- 10MiMo V2.6 Pro72.14%
- 11DeepSeek V4.1 Flash 101771.86%
- 12Kimi K371.72%
- 13Hy4 Preview71.70%
- 14GLM 5.3 Flash71.05%
- 15Hy470.55%
Source: Vals AI (Vals Legal Index) · Licence: CC BY 4.0 · Adaptation: Prudai selected the models, translated the explanations, added the gate verdict and the PrudentScore and rounded figures to two decimals.
| # Model | Score | Cost/task | Gate | PrudentScore |
|---|---|---|---|---|
| 1Wonder FullMeta | 77.86%± 1.82 | $3.72 | Rejected | no score |
| 2Gemini 4 ArgonGoogle | 76.78%± 1.84 | $6.82 | Passes | 79 |
| 3Gemini 4 CarbonGoogle | 75.93%± 1.85 | $7.20 | Passes | 77 |
| 4Muse EmberMeta | 75.69%± 1.84 | $0.07 | Rejected | no score |
| 5WonderMeta | 75.26%± 1.83 | $4.32 | Rejected | no score |
| 6GLM 5.3Z.ai | 74.62%± 1.84 | $2.68 | Rejected | no score |
| 7Inkling 2 SmallThinking Machines | 73.89%± 1.83 | $0.56 | Rejected | no score |
| 8Grok 4.7xAI (SpaceXAI) | 72.67%± 1.82 | $5.82 | Rejected | no score |
| 9Claude Opus 5.5Anthropic | 72.37%± 1.84 | $19.94 | Passes | 63 |
| 10MiMo V2.6 ProXiaomi | 72.14%± 1.83 | $0.18 | Rejected | no score |
| 11DeepSeek V4.1 Flash 1017DeepSeek | 71.86%± 1.82 | $0.59 | Rejected | no score |
| 12Kimi K3Moonshot AI | 71.72%± 1.80 | $3.47 | Rejected | no score |
| 13Hy4 PreviewTencent | 71.70%± 1.84 | $0.81 | Rejected | no score |
| 14GLM 5.3 FlashZ.ai | 71.05%± 1.83 | $0.23 | Rejected | no score |
| 15Hy4Tencent | 70.55%± 1.81 | $1.65 | Rejected | no score |
| 16Claude Haiku 5.5Anthropic | 70.29%± 1.83 | $2.36 | Passes | 59 |
| 17Step 5 PreviewStepFun | 70.16%± 1.78 | $0.84 | Rejected | no score |
| 18Gemini 3.8 FlashGoogle | 69.78%± 1.78 | $2.05 | Passes | 62 |
| 19Claude Fable 5.2Anthropic | 69.17%± 1.81 | $8.88 | Rejected | no score |
| 20DeepSeek V4 Pro (0813)DeepSeek | 69.01%± 1.77 | $0.87 | Rejected | no score |
| 21MiMo V2.6 FlashXiaomi | 68.99%± 1.77 | $0.07 | Rejected | no score |
| 22GPT-6.1 SolOpenAI | 68.72%± 1.77 | $2.75 | Passes | 59 |
| 23Muse Spark OpenMeta | 68.51%± 1.76 | $5.36 | Rejected | no score |
| 24GPT-6 AstraOpenAI | 68.39%± 1.74 | $13.94 | Passes | 56 |
| 25Claude Sonnet 5.5Anthropic | 68.02%± 1.81 | $13.37 | Passes | 62 |
| 26Mistral Large 4 (GA, 26.10)Mistral AI | 67.84%± 1.70 | $0.78 | Not yet assessed | no score |
| 27MiniMax-M3.1-FlashMiniMax | 67.07%± 1.78 | $0.41 | Rejected | no score |
| 28Kimi K3.1 ProMoonshot AI | 67.04%± 1.65 | $6.63 | Rejected | no score |
| 29Mistral Large 4 (preview)Mistral AI | 65.61%± 1.63 | $2.22 | Rejected | no score |
| 30Mercury 3Inception Labs | 62.96%± 1.62 | $0.10 | Not yet assessed | no score |
| 31GPT-6 LunaOpenAI | 62.20%± 1.57 | $0.31 | Passes | 54 |
| 32Beam 1.0Reflection AI | 61.80%± 1.61 | - | Rejected | no score |
| 33GPT-6.1 TerraOpenAI | 58.76%± 1.36 | $2.29 | Passes | 44 |
| 34Inkling SmallThinking Machines | 57.83%± 1.70 | $0.14 | Rejected | no score |
| 35Nemotron 3.5 Super VLNVIDIA | 55.37%± 1.25 | $0.16 | Rejected | no score |
| 36InklingThinking Machines | 54.16%± 1.66 | $0.70 | Rejected | no score |
Source: Vals AI (Vals Legal Index) · Licence: CC BY 4.0 · Adaptation: Prudai selected the models, translated the explanations, added the gate verdict and the PrudentScore and rounded figures to two decimals.
Vals AI · Version 1 · Measured 9 November 2026 · Go to the source. The PrudentScore is a composition by Prudai B.V. based on publicly published figures and does not represent the view or endorsement of Vals AI, Harvey or Artificial Analysis. Figures from Artificial Analysis appear in running text only, with attribution and a link, as that party's terms require [9].
Other benchmarks
- Legal Research Bench
- Harvey LAB (Vals implementation)
- LegalBench (Vals leaderboard)
- CaseLaw v2
- TaxEval v2
- Tax Agent Bench
- CorpFin v2
- Immigration Bench
- Access to Justice
- Public Benefits Bench
- Vals Index
- GPQA Diamond (Vals-run)
- MMLU Pro (Vals-run)
- Artificial Analysis Intelligence Index
Claude and Anthropic, GPT and OpenAI, Gemini and Google, Grok and xAI, Kimi and Moonshot AI, Muse and Meta, Mistral, DeepSeek, GLM and Z.ai, Harvey, Vals AI and Artificial Analysis are trademarks of their respective owners. They are named for identification only and this implies no partnership, sponsorship or endorsement.