WebsiteBenchWebsiteBench

Article

Rebuilt by Claude Opus 5: 273 of 1,759 hidden tests passed (15.5%)

reference.local/
Reference for Article, HomeReference
rebuild.local/
Rebuild by Claude Opus 5, HomeClaude Opus 5
    Every model on Article
    ModelPassed%Screenshot
    Claude Opus 527315.5
    GLM 5.3 Flash19911.3
    GPT 6.1 Sol18810.7
    DeepSeek V4.1 Flash1488.4
    Kimi K31257.1
    Claude Opus 5.5465.8
    Grok 4.6291.6
    Qwen 3.8 Max241.4