Atlas

Benchmarks

Access to Justice Benchmark

Vals

Vals AI's Access to Justice Benchmark v1, built with support from NextLadder Ventures, tests whether a model can be a reliable first point of contact for a person facing a civil legal problem without a lawyer. In each of 252 test scenarios that legal aid experts adapted from real legal-aid questions in six states, the model can search the web, read pages and question a GPT-5 simulated client for up to five exchanges, and a Decider v1.2 judge grades the conversation against the expert rubric (about 15 criteria a scenario, roughly a quarter of them facts the model must find out). The score is the share of rubric criteria met, pooled over all scenarios: a must-find-out criterion whose fact the opening question gave or the client volunteered is set aside, a scenario a provider's content filter blocked scores zero, and the reported uncertainty is criterion-level.

Models
21
Top score
74.9
Leader
Claude Sonnet 5.5
Updated
—
Resolution
~6 mo
Useful
Saturation
No trend
No estimate
Signal
29thpercentile
192 benchmarks from 0.8 to 3.5218 benchmarks from 3.5 to 6.1128 benchmarks from 6.1 to 8.895 benchmarks from 8.8 to 11.567 benchmarks from 11.5 to 14.128 benchmarks from 14.1 to 16.814 benchmarks from 16.8 to 19.49 benchmarks from 19.4 to 22.111 benchmarks from 22.1 to 24.72 benchmarks from 24.7 to 27.46 benchmarks from 27.4 to 30.13 benchmarks from 30.1 to 32.73 benchmarks from 32.7 to 35.45 benchmarks from 35.4 to 38.01 benchmark from 38.0 to 40.71 benchmark from 40.7 to 43.41 benchmark from 43.4 to 46.01 benchmark from 62.0 to 64.6

559th strongest of 785

Signal-to-noise +3.83

Difficulty
18thpercentile
1 benchmark from 6.0 to 14.81 benchmark from 14.8 to 23.72 benchmarks from 32.5 to 41.41 benchmark from 41.4 to 50.22 benchmarks from 50.2 to 59.08 benchmarks from 59.0 to 67.97 benchmarks from 67.9 to 76.719 benchmarks from 76.7 to 85.623 benchmarks from 85.6 to 94.441 benchmarks from 94.4 to 103.331 benchmarks from 103.3 to 112.141 benchmarks from 112.1 to 120.947 benchmarks from 120.9 to 129.872 benchmarks from 129.8 to 138.660 benchmarks from 138.6 to 147.5119 benchmarks from 147.5 to 156.3126 benchmarks from 156.3 to 165.2115 benchmarks from 165.2 to 174.037 benchmarks from 174.0 to 182.818 benchmarks from 182.8 to 191.79 benchmarks from 191.7 to 200.51 benchmark from 200.5 to 209.44 benchmarks from 209.4 to 218.2

641st hardest of 785

Capability needed ≲ 113.7

Organizations
21 dated models · 5 frontier modelsClosed weightsOpen weights
Data: Vals AIatlas.village.dev
#ModelSettingScoreSource
1Claude Sonnet 5.5max
74.9
Vals AI Access to Justice Benchmark leaderboard
2Muse Wonder Fullmax
74.7
Vals AI Access to Justice Benchmark leaderboard
3Claude Opus 5.5max
73.8
Vals AI Access to Justice Benchmark leaderboard
4GLM-5.3max
73.7
Vals AI Access to Justice Benchmark leaderboard
5Grok 4.7x-high
73.7
Vals AI Access to Justice Benchmark leaderboard
6MiniMax M3.1 Flashmax
73.3
Vals AI Access to Justice Benchmark leaderboard
7DeepSeek V4.1 Flash 1017max
73.1
Vals AI Access to Justice Benchmark leaderboard
8Muse Wondermax
73.0
Vals AI Access to Justice Benchmark leaderboard
9DeepSeek V4.1 Flashmax
72.9
Vals AI Access to Justice Benchmark leaderboard
10Muse Spark 1.3x-high
72.2
Vals AI Access to Justice Benchmark leaderboard
11Grok 4.6high
71.3
Vals AI Access to Justice Benchmark leaderboard
12GLM-5.3 Flashmax
71.1
Vals AI Access to Justice Benchmark leaderboard
13MiMo-V2.6-Flashnot reported
70.7
Vals AI Access to Justice Benchmark leaderboard
14MiMo-V2.6-Pronot reported
70.0
Vals AI Access to Justice Benchmark leaderboard
15MiniMax M3—
69.8
Vals AI Access to Justice Benchmark leaderboard
16Kimi K3.1 Promax
69.5
Vals AI Access to Justice Benchmark leaderboard
17Gemini 4 Argonhigh
68.5
Vals AI Access to Justice Benchmark leaderboard
18Kimi K3max
68.2
Vals AI Access to Justice Benchmark leaderboard
19Gemini 3.8 Flashhigh
67.1
Vals AI Access to Justice Benchmark leaderboard
20GPT-6.1 Solmax
65.8
Vals AI Access to Justice Benchmark leaderboard
21GPT-5.6 Lunamax
65.6
Vals AI Access to Justice Benchmark leaderboard
Loading Atlas data…