WebGameBench: leaderboard
Metric: Usable Rate (%; validly scored artifacts labelled Excellent or Usable). Source: arxiv.org. Saturation forecast: Around December 2026. 14 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4.7 | 76.9 |
| 2 | Claude Opus 4.6 | 73 |
| 3 | Gemini 3.1 Pro (Preview) | 63.6 |
| 4 | GPT-5.5 | 63.6 |
| 5 | Kimi K2.6 | 61 |
| 6 | DeepSeek V4 Flash (Non-reasoning) | 59.3 |
| 7 | GLM-5.1 | 52.8 |
| 8 | GLM-5 | 48.2 |
| 9 | Claude Sonnet 4.5 | 47.1 |
| 10 | DeepSeek V4 Pro (Non-reasoning) | 44.1 |
| 11 | Kimi K2.5 | 38.3 |
Interactive version: theaggregate.ai/benchmark?slug=webgamebench · How It Works · Data refreshed daily, snapshot 2026-09-25.