Benchmarks · Vals AI
Vals Index
General agent index across finance, coding, legal and tax, weighted by each sector's share of US GDP. Version 2.2 is current; 2.1 is archived.
- Models measured
- 50
- Top
- 68.90 %Gemini 4 Argonwithin the standard error of ±0.97
- Version
- 2.1
- Measured
- 18 October 2026
01
Ranking
At the top: Top 15
Vals AI Legal Research Bench, all-pass [1]. Hatched bars are rejected by the compliance gate.
- 1Gemini 4 Argon68.90%
- 2Claude Sonnet 5.567.04%
- 3Claude Opus 5.566.97%
- 4Claude Fable 5.165.83%
- 5Wonder Full64.57%
- 6Claude Opus 563.67%
- 7GPT-6 Astra63.13%
- 8Claude Fable 561.39%
- 9GPT-6.1 Sol61.15%
- 10Wonder61.06%
- 11Muse Spark 1.3 (Max)58.16%
- 12GPT-5.6 Sol58.01%
- 13GPT-6 Sol57.54%
- 14DeepSeek V4.1 Flash 101756.58%
- 15MiMo V2.6 Pro55.20%
Source: Vals AI (Vals 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 |
|---|---|---|---|---|
| 1Gemini 4 ArgonGoogle | 68.90%± 0.97 | $15.68 | Passes | 79 |
| 2Claude Sonnet 5.5Anthropic | 67.04%± 0.92 | $21.34 | Passes | 62 |
| 3Claude Opus 5.5Anthropic | 66.97%± 0.89 | $32.14 | Passes | 63 |
| 4Claude Fable 5.1Anthropic | 65.83%± 1.11 | $28.71 | Rejected | no score |
| 5Wonder FullMeta | 64.57%± 1.07 | $8.56 | Rejected | no score |
| 6Claude Opus 5Anthropic | 63.67%± 0.96 | $19.31 | Passes | 72 |
| 7GPT-6 AstraOpenAI | 63.13%± 1.19 | $18.46 | Passes | 56 |
| 8Claude Fable 5Anthropic | 61.39%± 1.00 | $29.59 | Rejected | no score |
| 9GPT-6.1 SolOpenAI | 61.15%± 1.01 | $3.24 | Passes | 59 |
| 10WonderMeta | 61.06%± 1.07 | $11.44 | Rejected | no score |
| 11Muse Spark 1.3 (Max)Meta | 58.16%± 1.19 | $3.79 | Rejected | no score |
| 12GPT-5.6 SolOpenAI | 58.01%± 1.02 | $14.24 | Passes | 57 |
| 13GPT-6 SolOpenAI | 57.54%± 1.01 | $7.58 | Passes | 45 |
| 14DeepSeek V4.1 Flash 1017DeepSeek | 56.58%± 1.27 | $0.64 | Rejected | no score |
| 15MiMo V2.6 ProXiaomi | 55.20%± 1.18 | $0.41 | Rejected | no score |
| 16Claude Opus 4.8Anthropic | 55.10%± 1.01 | $13.14 | Passes | 62 |
| 17Grok 4.7xAI (SpaceXAI) | 54.95%± 1.07 | $12.11 | Rejected | no score |
| 18Gemini 3.8 FlashGoogle | 54.89%± 1.02 | $5.71 | Passes | 62 |
| 19Claude Haiku 5.5Anthropic | 54.44%± 1.41 | $3.00 | Passes | 59 |
| 20GLM 5.3Z.ai | 53.51%± 1.30 | $7.25 | Rejected | no score |
| 21MiMo V2.6 FlashXiaomi | 53.23%± 1.10 | $0.20 | Rejected | no score |
| 22Muse Spark 1.3Meta | 53.20%± 1.07 | $3.38 | Rejected | no score |
| 23GPT-5.6 TerraOpenAI | 53.09%± 1.29 | $5.78 | Passes | 53 |
| 24Grok 4.6xAI (SpaceXAI) | 52.09%± 1.15 | $4.49 | Rejected | no score |
| 25Claude Sonnet 5Anthropic | 51.90%± 1.10 | $13.69 | Passes | 59 |
| 26GPT-5.6 LunaOpenAI | 51.69%± 1.06 | $0.82 | Passes | 56 |
| 27DeepSeek V4.1 FlashDeepSeek | 51.32%± 1.13 | $0.33 | Rejected | no score |
| 28Gemini 3.7 FlashGoogle | 51.27%± 1.10 | $4.67 | Passes | 58 |
| 29GPT-6 LunaOpenAI | 51.22%± 1.07 | $0.43 | Passes | 54 |
| 30Ember-1Fireworks AI | 50.94%± 1.30 | $9.66 | Not yet assessed | no score |
| 31Kimi K3Moonshot AI | 50.30%± 0.99 | $6.44 | Rejected | no score |
| 32Hy4 PreviewTencent | 49.94%± 1.15 | $1.44 | Rejected | no score |
| 33Step 5 PreviewStepFun | 49.35%± 1.22 | $2.63 | Rejected | no score |
| 34Muse Spark 1.2Meta | 49.29%± 1.10 | $1.94 | Rejected | no score |
| 35Mistral Large 4 (preview)Mistral AI | 48.66%± 1.13 | $14.07 | Rejected | no score |
| 36Qwen 3.8 MaxAlibaba | 48.27%± 1.34 | $4.96 | Not yet assessed | no score |
| 37DeepSeek V4 Flash 0731DeepSeek | 47.96%± 1.24 | $0.85 | Rejected | no score |
| 38DeepSeek V4 Pro (0813)DeepSeek | 47.63%± 1.12 | $3.35 | Rejected | no score |
| 39Muse Spark OpenMeta | 46.61%± 1.20 | $4.08 | Rejected | no score |
| 40Gemini 3.5 FlashGoogle | 44.79%± 1.10 | $3.30 | Passes | 47 |
| 41Grok 4.5xAI (SpaceXAI) | 44.72%± 1.17 | $2.42 | Rejected | no score |
| 42GLM 5.3 FlashZ.ai | 44.58%± 1.41 | $1.03 | Rejected | no score |
| 43DeepSeek V4 Pro (2026-04-23)DeepSeek | 38.63%± 1.16 | $1.31 | Rejected | no score |
| 44MiniMax-M3MiniMax | 36.53%± 1.17 | $3.08 | Rejected | no score |
| 45MiMo V2.5 ProXiaomi | 33.84%± 1.17 | $0.16 | Rejected | no score |
| 46Gemini 3.1 Pro Preview (02/26)Google | 33.44%± 1.13 | $2.18 | Rejected | no score |
| 47GPT 5.4 MiniOpenAI | 33.17%± 1.28 | $1.45 | Passes | 25 |
| 48InklingThinking Machines | 28.67%± 1.10 | $1.46 | Rejected | no score |
| 49Inkling SmallThinking Machines | 25.46%± 1.03 | $0.28 | Rejected | no score |
| 50Mercury 2.5Inception Labs | 8.86%± 0.45 | $0.29 | Not yet assessed | no score |
Source: Vals AI (Vals 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 2.1 · Measured 18 October 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 Legal 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.