List your company

Alignment on Arena-Hard (hard prompt gemini)

70.4Hard Prompt Gemini ScoreSnapMLA
54.90458.92762.9566.973Feb 11, 2026

Evaluation results

MethodLinks
SnapMLABackbone=LongCat-Flash-thinking, Precision=FP82026.0270.4
FlashMLABackbone=LongCat-Flash-thinking, Precision=BF162026.0269.9
FlashMLABackbone=DeepSeek-V3.1, Precision=BF162026.0257.1
SnapMLABackbone=DeepSeek-V3.1, Precision=FP82026.0255.5