WebVR: leaderboard
Metric: Overall rubric score (0-100): share of rubric criteria satisfied on WebVR's 175 webpage-recreation tasks: the model watches a screen-recording video of a synthesized webpage (plus its image assets) and writes one standalone HTML file; the page is rendered and replayed under a fixed interaction script and a Kimi-K2.5 judge checks it against the task's atomic visual rubric (32-248 binary criteria) using the code and the rendered video; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 19 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Kimi K2.5 | 79.14 | #139 |
| 2 | Claude Sonnet 4.6 | 78.49 | #85 |
| 3 | GPT-5.2 (Thinking) | 77.93 | #105 (GPT-5.2) |
| 4 | Claude Opus 4.6 | 77.33 | #60 |
| 5 | Gemini 3.1 Pro (Preview) | 76.69 | #54 |
| 6 | Seed 2.0 Pro | 71.88 | #111 |
| 7 | Gemini 3 Flash | 69.49 | #93 |
| 8 | Gemini 3 Pro | 67.32 | #77 |
| 9 | Gemini 2.5 Pro | 63.09 | #145 |
| 10 | Seed 1.8 | 61.98 | #136 |
| 11 | Qwen 3.5 397B A17B | 61.33 | #141 |
| 12 | Claude 3.7 Sonnet | 61.21 | #241 |
| 13 | Gemini 2.5 Flash | 55.62 | #237 |
| 14 | Qwen 3 VL 235B A22B (Thinking) | 46.8 | #228 (Qwen 3 VL 235B A22B) |
| 15 | GPT-4.1 | 45.85 | #240 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=webvr · How It Works · Data refreshed daily, snapshot 2026-10-11.