Atlas

Benchmarks

← All benchmarks

HLAB-Vals-1.1 - Criteria Pass Rate - Tax

Professional Work · 2026-10-19

Criteria pass rate (judge-averaged, Harvey's definition) on the Tax practice-area tasks of HLAB-Vals-1.1, distinct from that area's task pass rate.

Top models (higher is better)

ModelScore
Muse Spark Open97.0
Gemini 4 Argon94.8
Muse Wonder Full92.6
Muse Wonder92.5
MiMo-V2.6-Pro91.5
GLM-5.391.1
Kimi K390.3
MiMo-V2.6-Flash89.6
Gemini 3.8 Flash89.5
Mistral Large 4 Preview89.5
GLM-5.3 Flash88.6
Beam 1.0 501B A23B87.8
DeepSeek V4.1 Flash 101787.8
GPT-6.1 Sol86.0
DeepSeek V4 Pro 081385.6
Claude Haiku 5.585.3
GPT-6 Astra84.8
Grok 4.784.6
MiniMax M3.1 Flash84.3
Mercury 382.6
Claude Opus 5.581.8
Inkling-Small80.4
GPT-6 Luna79.4
Hy4 Preview64.0
Claude Sonnet 5.562.7
Inkling51.0
Loading Atlas data…