LoHoSearch: leaderboard
LoHoSearch evaluates long-horizon search agents on 544 human-verified questions requiring multi-step open-web research across 11 topical domains; this board reports the paper's end-to-end answer accuracy.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Around October 2027. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 | 34.74 | #31 |
| 2 | DeepSeek V4 Pro | 15.99 | #102 |
| 3 | Claude Opus 4.6 | 15.62 | #63 |
| 4 | Kimi K2.6 | 15.53 | #104 |
| 5 | Gemini 3.1 Pro (Preview) | 13.32 | #58 |
| 6 | GLM-5.1 | 12.77 | #97 |
| 7 | Claude Opus 4.7 | 10.29 | #49 |
| 8 | DeepSeek V4 Flash | 10.02 | #137 |
| 9 | LongCat-Flash-Thinking-2601 | 9.74 | #229 |
| 10 | MiniMax-M2.7 | 2.48 | #258 |
| 11 | MiniMax-M2.5 | 2.29 | #303 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=lohosearch · How It Works · Data refreshed daily, snapshot 2026-10-20.