OpenRouter HLE (Search): leaderboard
OpenRouter's Humanity's Last Exam run with web search: 2,158 expert-written questions answered through a search provider (Exa, Perplexity, Parallel, or the OpenAI/Anthropic native tools) under a 1, 5, or 25-turn budget. The board varies the search stack rather than the model, so each model is scored on its best configuration.
Metric: Best-configuration Accuracy (%). Source: openrouter.ai. Saturation forecast: Around April 2027. 2 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5 (High) | 77.4 | #16 (Claude Opus 5) |
| 2 | GPT-5.6 Sol (High) | 71.1 | #22 (GPT-5.6 Sol) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=openrouter-hle-search · How It Works · Data refreshed daily, snapshot 2026-11-05.