CUJBench (Browser-Only Agent): leaderboard
Metric: Root-cause accuracy: single-prediction root-cause accuracy (A@1, %: component and fault type both correct; a run that never submits counts as wrong) on the 25-scenario CUJBench evaluation subset of browser-to-backend failure diagnoses (OpenTelemetry Demo and Tractor Store applications), one run per scenario; ReAct tool loop capped at 15 turns over the 7 browser-visible tools (screenshots, network, console, DOM, frontend spans); higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 6 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 52 |
| 2 | Claude Sonnet 4.6 | 52 |
| 3 | GLM-4.6V | 16 |
Interactive version: theaggregate.ai/benchmark?slug=cujbench-browser-only-agent · How It Works · Data refreshed daily, snapshot 2026-10-07.