model,harness,provider,mean_pass_rate,ci_low,ci_high,n_evaluations
anthropic/claude-opus-4-8,pi,anthropic,50.2,40.2,60.2,83
openai/gpt-5.5,openai-codex,openai,50.2,40.8,59.5,93
anthropic/claude-opus-4-7,pi,anthropic,49.6,40.0,59.1,78
anthropic/claude-sonnet-4-6,pi,anthropic,48.6,39.0,58.2,83
gemini/gemini-3.5-flash,pi,google,47.1,37.5,56.7,92
anthropic/claude-opus-4-6,claude-code,anthropic,46.7,36.4,56.9,80
anthropic/claude-opus-4-6,pi,anthropic,45.7,35.9,55.4,81
gemini/gemini-3.1-pro-preview,pi,google,45.3,36.7,53.8,95
openai/gpt-5.5,pi,openai,44.8,35.0,54.6,77
anthropic/claude-opus-4-7,claude-code,anthropic,44.2,33.7,54.6,77
anthropic/claude-sonnet-4-6,claude-code,anthropic,43.6,33.9,53.2,83
openai/gpt-5.4,openai-codex,openai,41.3,32.5,50.1,92
anthropic/claude-opus-4-8,claude-code,anthropic,39.5,29.5,49.4,76
openai/gpt-5.4,pi,openai,38.5,29.3,47.6,81
xai/grok-4.3,pi,xai,16.0,10.1,21.9,100
xai/grok-4.20-beta-0309-reasoning,pi,xai,13.7,8.1,19.2,100