Atlas

Benchmarks

← All benchmarks

Vals Legal Index

Indexes · 2026-10-23

Vals Legal Index v1 weights nine legal practice-area lines by their share of 2022 U.S. law-office revenue (Census NAICS 541110). Each area is scored by the Vals benchmarks that cover it, averaging Legal Research Bench all-pass accuracy and HLAB-Vals-1.1 criteria pass rate where both apply, and the index is 0.978 x the revenue-weighted areas plus 0.022 x Tax Agent Bench's weighted pass rate. Component runs match the standalone Vals leaderboards, which Atlas imports separately. Tracked for comparison only; as an index it is not a capability-model input.

Top models (higher is better)

ModelScore
Muse Wonder Full77.9
Gemini 4 Argon76.7
Muse Wonder75.3
GLM-5.374.6
Qwen3.8 Max (0902)73.9
Grok 4.772.6
MiMo-V2.6-Pro72.1
DeepSeek V4.1 Flash 101771.9
Kimi K371.7
Hy4 Preview71.7
GLM-5.3 Flash71.1
Claude Haiku 5.570.3
Gemini 3.8 Flash69.8
DeepSeek V4 Pro 081369.0
MiMo-V2.6-Flash69.0
GPT-6.1 Sol68.7
Muse Spark Open68.5
GPT-6 Astra68.4
MiniMax M3.1 Flash67.1
Mistral Large 4 Preview65.6
Mercury 363.0
GPT-6 Luna62.2
Beam 1.0 501B A23B61.8
Inkling-Small57.8
Inkling54.2
Loading Atlas data…