WebsiteBenchWebsiteBench

Bluemercury

Rebuilt by GPT 6.1 Sol: 414 of 694 hidden tests passed (59.7%)

reference.local/
Reference for Bluemercury, HomeReference
rebuild.local/
Rebuild by GPT 6.1 Sol, HomeGPT 6.1 Sol
    Every model on Bluemercury
    ModelPassed%Screenshot
    GPT 6.1 Sol41459.7
    Claude Opus 5.540257.9
    Grok 4.640157.8
    Claude Opus 539256.5
    DeepSeek V4.1 Flash36552.6
    GLM 5.3 Flash29342.2
    Kimi K3537.6
    Qwen 3.8 Max253.6

    Every result in the WaffleRun the benchmark