OpenRouter BrowseComp (Search): leaderboard
OpenRouter's BrowseComp run: 1,266 questions built to be unfindable in a single search, rewarding persistent multi-step browsing. Scored across search providers and turn budgets; each model takes its best configuration.
Metric: Best-configuration Accuracy (%). Source: openrouter.ai. Saturation forecast: Around February 2027. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5 (High) | 89 | #16 (Claude Opus 5) |
| 2 | GPT-5.6 Sol (High) | 82.4 | #22 (GPT-5.6 Sol) |
| 3 | DeepSeek V4 Flash (0731) (High) | 77 | #84 (DeepSeek V4 Flash (0731)) |
| 4 | GPT-5.6 Luna (xHigh) | 74 | #79 (GPT-5.6 Luna) |
| 5 | Claude Opus 5 (Medium) | 12 | #16 (Claude Opus 5) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=openrouter-browsecomp-search · How It Works · Data refreshed daily, snapshot 2026-11-05.