ReactBench v1 Write React
Code · 2026-07-14
The 27-task Write React subset of ReactBench v1 asks coding agents to implement real features or fixes from open-source pull requests. A rollout passes only when held-out behavioral tests succeed and the changed surface introduces no new graded React Doctor issue.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Fable 5 | 65.9 |
| GPT-5.6 Terra | 64.4 |
| Claude Opus 5 | 63.7 |
| GPT-5.6 Sol | 57.8 |
| GPT-5.6 Luna | 54.1 |
| Opus 4.8 | 50.4 |
| Grok 4.5 | 48.1 |
| GLM-5.2 | 47.4 |
| Kimi K3 | 45.2 |
| Sonnet 5 | 42.2 |
| Muse Spark 1.1 | 35.6 |
| Gemini 3.5 Flash | 33.3 |
| Gemini 3.1 Pro Preview | 32.6 |
| Kimi K2.7 Code | 31.9 |
| Composer 2.5 | 27.4 |
| Inkling | 14.1 |