Skip to content

Benchmarks · Vals AI

Immigration Bench

Measures 262 expert-written questions on US immigration law (seven personas), graded per question rubric; split into all-pass, intake checks and content checks.

Models measured
26
Top
75.44 %Claude Fable 5.1within the standard error of ±1.40
Version
1
Measured
9 November 2026
Go to the sourceThis benchmark does not count towards the PrudentScore.

01

Ranking

At the top: Top 15

Vals AI Legal Research Bench, all-pass [1]. Hatched bars are rejected by the compliance gate.

  1. 1Claude Fable 5.175.44%
  2. 2Claude Fable 5.273.98%
  3. 3Claude Opus 5.572.14%
  4. 4Wonder Full71.22%
  5. 5Qwen 3.8 Max70.14%
  6. 6Wonder69.62%
  7. 7Claude Sonnet 5.568.96%
  8. 8Kimi K368.26%
  9. 9Gemini 4 Argon68.08%
  10. 10Grok 4.768.05%
  11. 11GPT-6.1 Sol67.81%
  12. 12GPT-6 Astra67.69%
  13. 13MiMo V2.6 Pro67.50%
  14. 14GLM 5.366.73%
  15. 15Muse Spark 1.266.54%

Source: Vals AI (Immigration Bench) · 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.

Passes the compliance gateRejected by the compliance gate
Immigration Bench 1
# ModelScoreCost/taskGatePrudentScore
1Claude Fable 5.1Anthropic75.44%± 1.40$2.11Rejectedno score
2Claude Fable 5.2Anthropic73.98%± 1.33$0.64Rejectedno score
3Claude Opus 5.5Anthropic72.14%± 1.45$2.16Passes63
4Wonder FullMeta71.22%± 1.46$0.39Rejectedno score
5Qwen 3.8 MaxAlibaba70.14%± 1.57$0.12Not yet assessedno score
6WonderMeta69.62%± 1.44$0.30Rejectedno score
7Claude Sonnet 5.5Anthropic68.96%± 1.54$1.16Passes62
8Kimi K3Moonshot AI68.26%± 1.51$0.18Rejectedno score
9Gemini 4 ArgonGoogle68.08%± 1.45$0.51Passes79
10Grok 4.7xAI (SpaceXAI)68.05%± 1.60$0.24Rejectedno score
11GPT-6.1 SolOpenAI67.81%± 1.53$0.23Passes59
12GPT-6 AstraOpenAI67.69%± 1.55$1.37Passes56
13MiMo V2.6 ProXiaomi67.50%± 1.55$0.02Rejectedno score
14GLM 5.3Z.ai66.73%± 1.62$0.12Rejectedno score
15Muse Spark 1.2Meta66.54%± 1.59$0.04Rejectedno score
16Grok 4.6xAI (SpaceXAI)65.98%± 1.53$0.10Rejectedno score
17DeepSeek V4.1 FlashDeepSeek65.64%± 1.64$0.05Rejectedno score
18Muse Spark 1.3Meta65.00%± 1.56$0.05Rejectedno score
19Muse Spark 1.3 (Max)Meta64.93%± 1.56$0.06Rejectedno score
20Mistral Large 4 (preview)Mistral AI63.71%± 1.66$0.15Rejectedno score
21Gemini 3.8 FlashGoogle63.41%± 1.61$0.17Passes62
22GPT-6 LunaOpenAI62.53%± 1.57$0.01Passes54
23Claude Haiku 5.5Anthropic61.79%± 1.64$0.09Passes59
24DeepSeek V4 Pro (0813)DeepSeek61.66%± 1.68$0.05Rejectedno score
25MiniMax-M3MiniMax57.53%± 1.62$0.04Rejectedno score
26Gemini 3.5 Flash LiteGoogle55.66%± 1.70$0.02Passes38

Source: Vals AI (Immigration Bench) · 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

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.