Meituan LongCat-2.0 Card - RWSearch: leaderboard
RWSearch, published by AGI-Eval, gives search agents 200 expert-written questions in a Chinese-language context, each with one objective answer that needs real-time web retrieval, long reasoning chains and numerical comparison. Meituan's LongCat-2.0 model card measured every model in-house under one unified harness; score in percent, higher is better.
Metric: Score (%). Source: huggingface.co. Saturation forecast: Around February 2027. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 | 85.3 | #33 |
| 2 | Claude Opus 4.6 | 81.3 | #67 |
| 3 | Claude Opus 4.7 | 79.3 | #51 |
| 4 | LongCat 2.0 | 78.8 | #220 |
| 5 | Claude Opus 4.8 | 77.3 | #39 |
| 6 | Gemini 3.1 Pro (Preview) | 76.3 | #64 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=meituan-longcat-2-0-card-rwsearch · How It Works · Data refreshed daily, snapshot 2026-11-05.