Atlas

Benchmarks

← All benchmarks

Immigration Bench - Content Checks

Professional Work · 2026-10-28

Vals' Content Checks panel of Immigration Bench, apparently the rubric items other than intake checks. The page does not define the panel or its denominator, and its standard errors are zero placeholders. Diagnostic only.

Top models (higher is better)

ModelScore
Claude Fable 5.173.2
Claude Opus 5.570.2
Muse Wonder Full68.7
Muse Wonder67.4
Grok 4.766.7
Claude Sonnet 5.566.2
MiMo-V2.6-Pro65.7
Kimi K365.6
Gemini 4 Argon65.6
GPT-6.1 Sol65.1
GPT-6 Astra64.9
GLM-5.364.5
Muse Spark 1.264.3
DeepSeek V4.1 Flash64.2
Grok 4.663.3
Muse Spark 1.362.3
Gemini 3.8 Flash61.2
GPT-6 Luna60.8
Mistral Large 4 Preview60.2
DeepSeek V4 Pro 081359.1
Claude Haiku 5.557.6
MiniMax M356.2
Gemini 3.5 Flash-Lite51.7
Loading Atlas data…