Benchmark Explorer

Explore how models perform on various benchmarks

Benchmarks

🔍

Vals Index

A composite benchmark aggregating performance across vals.ai's enterprise-focused evaluations spanning legal, finance, healthcare, and coding domains. The index provides a single summary score reflecting a model's overall capability across the vals.ai benchmark suite, weighting performance across diverse professional and technical tasks.
Source:

Model Performance

#1
Claude Sonnet 4.6
67.7%
#2
Claude Opus 4.6 (Thinking)
65.9%
#3
GPT-5.2
64.5%
#4
Claude Opus 4.6
64.2%
#5
Claude Opus 4.5 (Thinking)
63.8%
#6
Claude Opus 4.5
61.3%
#7
Grok 4.20 (Reasoning)
51.0%
#8
DeepSeek V3.2
49.4%
#9
Gemini 2.5 Pro (Jun 2025)
49.1%
#10
DeepSeek V3.2 (Thinking)
47.1%
#11
Grok 4.20
39.8%
#12
GPT OSS 120B
39.3%