DrivingBench (First Attempt): leaderboard
DrivingBench's cone-course progress on each model's first attempt only, before any in-chat reflection on a failed run; share of the course centerline reached while staying within 4 m of it.
Metric: Course progress (%, first attempt). Source: drivingbench.com. Saturation forecast: Around May 2027. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-6 Astra (Medium) | 48.82 | #6 (GPT-6 Astra) |
| 2 | Claude Fable 5.1 (Medium) | 9.45 | #14 (Claude Fable 5.1) |
| 3 | Grok 4.6 (Medium) | 7.87 | #44 (Grok 4.6) |
| 4 | GPT-5.6 Sol (Medium) | 6.3 | #29 (GPT-5.6 Sol) |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=drivingbench-first-attempt · How It Works · Data refreshed daily, snapshot 2026-11-12.