RPCBench - Proactive Detection: leaderboard
Metric: Detection Rate (%). Source: arxiv.org. Saturation forecast: Around December 2026. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 60 |
| 2 | Qwen 3.5 122B A10B | 59.8 |
| 3 | GPT-5.5 | 59.6 |
| 4 | DeepSeek V4 Pro | 59.3 |
| 5 | Qwen 3.5 Plus | 59.1 |
| 6 | Qwen 3.5 397B A17B | 59 |
| 7 | Claude Sonnet 4.6 | 58 |
| 8 | Qwen 3.5 35B A3B | 56.5 |
| 9 | DeepSeek V4 Flash | 53.8 |
| 10 | Llama 3.1 70B Instruct | 22.7 |
| 11 | Llama 3.1 8B Instruct | 19.1 |
Interactive version: theaggregate.ai/benchmark?slug=rpcbench-proactive-detection · How It Works · Data refreshed daily, snapshot 2026-09-24.