BridgeBench Overall: leaderboard
BridgeBench V3 coding leaderboard: each model is rated 0-1000 on nine axes (reasoning, front end, back end, one-shot, security, trust, efficiency, speed, cost) and ranked by Overall, the mean of the nine; higher is better on every axis. Speed and cost are not capability measures, so cheap, fast models gain on Overall. It replaced the head-to-head arena ladder in September 2026 and moved from a 1-10 to a 0-1000 scale on 2026-09-24.
Metric: Overall (mean of nine axes, 0-1000). Source: www.bridgebench.ai. Saturation forecast: Around September 2027. 30 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5.5 | 760 | #3 |
| 2 | GPT-6 Astra | 728 | #5 |
| 3 | Claude Fable 5.1 | 689 | #10 |
| 4 | Claude Sonnet 5.5 | 647 | #13 |
| 5 | GPT-6.1 Sol | 646 | #6 |
| 6 | GPT-6 Sol | 643 | #17 |
| 7 | Claude Fable 5 | 628 | #19 |
| 8 | Gemini 4 Argon | 623 | #9 |
| 9 | Muse Wonder | 611 | #16 |
| 10 | Muse Spark 1.3 | 606 | #31 |
| 11 | Claude Haiku 5.5 | 601 | #46 |
| 12 | GPT-5.6 Sol | 600 | #22 |
| 13 | Gemini 3.8 Flash | 584 | #30 |
| 14 | Step 5 Preview | 572 | #54 |
| 15 | MiMo-V2.6-Pro | 568 | #51 |
Interactive version: theaggregate.ai/benchmark?slug=bridgebench-overall · How It Works · Data refreshed daily, snapshot 2026-11-07.