Benchmarks
326 sources · 1,256 models × 2,956 benchmarks · 154,484 scoresStrict completion on Co-Design Bench's 173 Repository Editing tasks: coordinated changes to components, routes, state, data flow and content in runnable multi-file frontend repositories. Experts split compound requests into 1,778 separately auditable criteria, about 10.3 per task, and 166 tasks need two or more edit types.
547th strongest of 779 fitted benchmarks
Signal-to-noise +3.91 for the Co-Design Bench family
Another benchmark represents this family in the fit.
| # | Model | Setting | Score | Source |
|---|---|---|---|---|
| 1 | GPT-6 Astra | not reported · Claude Code | 66.7 | |
| 2 | Qwen3.8 27B | not reported · Claude Code | 64.5 | |
| 3 | Qwen3.8 Max (0902) | not reported · Claude Code | 64.4 | |
| 4 | Claude Opus 5 | not reported · Claude Code | 64.2 | |
| 5 | GPT-5.6 Sol | not reported · Claude Code | 62.7 | |
| 6 | Kimi K3 | not reported · Claude Code | 61.0 | |
| 7 | GLM-5.3 Flash | not reported · Claude Code | 59.4 | |
| 8 | Opus 4.8 | not reported · Claude Code | 57.5 | |
| 9 | GLM-5.2 | not reported · Claude Code | 56.4 | |
| 10 | Grok 4.6 | not reported · Claude Code | 53.6 |