Atlas

Benchmarks

← All benchmarks

Legal Research Bench Weighted Pass

Professional Work · 2026-06-23

This row reports the practice-area score for Weighted Pass tasks in Vals AI Legal Research Bench.

Top models (higher is better)

ModelScore
Gemini 4 Carbon91.6
Gemini 4 Argon91.0
Claude Opus 590.6
Muse Wonder Full89.8
Grok 4.688.3
GPT-5.6 Sol87.8
GLM-5.387.7
Muse Wonder87.7
GLM-5.3 Flash87.5
Gemini 3.8 Flash87.5
Muse Spark 1.386.8
Grok 4.786.6
DeepSeek V4.1 Flash 101786.3
Qwen3.8-Max86.1
Opus 4.886.1
Ember-186.0
Muse Spark Open86.0
MiMo-V2.6-Pro85.5
Sonnet 4.685.3
Mistral Large 4 Preview85.3
Muse Spark 1.285.3
Hy4 Preview85.2
GPT-5.6 Terra85.1
GPT-5.6 Luna85.0
GPT-6 Astra84.8
Claude Haiku 5.584.8
Step 5 Preview84.8
Kimi K384.6
Muse Spark 1.184.6
Sonnet 584.6
DeepSeek V4.1 Flash84.5
Grok 4.584.4
GPT-6.1 Sol84.3
Gemini 3.7 Flash84.1
Opus 4.784.1
GPT-5.584.0
Gemini 3.5 Flash83.9
DeepSeek V4 Pro 081383.8
MiniMax M3.1 Flash83.1
GLM-5.282.9
Loading Atlas data…