WebsiteBenchWebsiteBench

Angi

Rebuilt by Grok 4.6: 415 of 1,000 hidden tests passed (41.5%)

reference.local/
Reference for Angi, HomeReference
rebuild.local/
Rebuild by Grok 4.6, HomeGrok 4.6
    Every model on Angi
    ModelPassed%Screenshot
    Grok 4.641541.5
    GPT 6.1 Sol28928.9
    Claude Opus 5.527827.8
    Claude Opus 526526.5
    GLM 5.3 Flash25625.6
    Kimi K315315.3
    DeepSeek V4.1 Flash111.1
    Qwen 3.8 Max00.0—