OmniGameArena - ObstacleRun2D: leaderboard
The "ObstacleRun2D" subtask of OmniGameArena. OmniGameArena (arXiv) — Unreal-Engine-5 game environments testing vision-language agents on real-time perception, control and planning across solo and co-op games; normalized game score (%).
Metric: Normalized game score (%). Source: arxiv.org. Saturation forecast: Around July 2027. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 | 47.3 | #31 |
| 2 | Claude Opus 4.6 | 33.8 | #63 |
| 3 | Gemini 3.1 Flash Lite (Preview) | 27.8 | #194 |
| 4 | Claude Opus 4.7 | 22 | #49 |
| 5 | GPT-5.4 | 12.2 | #79 |
| 6 | Qwen 3.5 397B A17B | 11.4 | #146 |
| 7 | Kimi K2.5 | 10.9 | #144 |
| 8 | Gemini 3.1 Pro (Preview) | 10.2 | #58 |
| 9 | Qwen 3.5 122B A10B | 9.2 | #173 |
| 10 | Claude Sonnet 4.6 | 7.5 | #90 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=omnigamearena-obstaclerun2d · How It Works · Data refreshed daily, snapshot 2026-10-20.