GenXR-Bench - Bug Repair: leaderboard
The 21 GenXR-Bench tasks that start from an XR Blocks SDK snapshot with a reproduced behavioural defect: the agent must find and fix it; the score is the share passed in one attempt. Most models pass over 80%, far more than on the feature and app tasks.
Metric: Pass@1 on the 21 SDK bug-repair tasks (%). Source: arxiv.org. Saturation forecast: Estimated already saturated. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.6 Sol (xHigh) | 95.2 | #30 (GPT-5.6 Sol) |
| 2 | Gemini 3.8 Flash (High) | 90.5 | #36 (Gemini 3.8 Flash) |
| 3 | Gemini 3.7 Flash (High) | 90.5 | #32 (Gemini 3.7 Flash) |
| 4 | Gemini 3.5 Flash (High) | 85.7 | #101 (Gemini 3.5 Flash) |
| 5 | Gemini 3.6 Flash (High) | 85.7 | #76 (Gemini 3.6 Flash) |
| 6 | GPT-5.6 Luna (xHigh) | 85.7 | #121 (GPT-5.6 Luna) |
| 7 | Claude Opus 5 (Medium) | 85.7 | #22 (Claude Opus 5) |
| 8 | Gemini 3.1 Pro (Preview) (High) | 81 | #94 (Gemini 3.1 Pro (Preview)) |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=genxr-bench-bug-repair · How It Works · Data refreshed daily, snapshot 2026-11-13.