ComponentBench (Browser Use): leaderboard
ComponentBench (2026) gives computer-use agents short, programmatically verified tasks on modern web UI components, such as setting a slider, picking a date or reordering a drag-and-drop list, 2,910 tasks over 97 component types with a 20-step budget. In this setting the agent runs in the tool-rich browser-use framework with DOM access. Score is the task pass rate in percent; higher is better.
Metric: Pass rate (%; share of tasks whose programmatic verifier passes; the browser-use agent framework with screenshots, serialized DOM and grounded element references, up to 4 actions per step; ComponentBench-Full: 2,910 programmatically verified tasks on 97 canonical UI components across widely used web component libraries; 20-step budget; one deterministic run). Source: arxiv.org. Saturation forecast: Estimated already saturated. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash | 95.2 | #164 |
| 2 | GPT-5.4 | 90.4 | #123 |
| 3 | Gemini 3.1 Flash Lite | 87.4 | #364 |
| 4 | GPT-5 Mini | 87 | #348 |
| 5 | GPT-5.4 Mini | 85.8 | #389 |
| 6 | Qwen 3 VL 235B A22B | 78.8 | #479 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=componentbench-browser-use · How It Works · Data refreshed daily, snapshot 2026-11-13.