List your company

Chat Performance on Arena-Hard-Auto

92.8ScoreGR3
65.55272.62679.786.774Mar 11, 2026

Evaluation results

MethodLinks
GR3Model Backbone=Qwen3–8B2026.0392.81,178
GRPOModel Backbone=Qwen3–8B2026.0390.62,343
GR3Model Backbone=Qwen3–4B2026.0385.91,377
GRPOModel Backbone=Qwen3–4B2026.0385.82,374
InitialModel Backbone=Qwen3–8B2026.0377.21,171
InitialModel Backbone=Qwen3–4B2026.0366.61,139