RPCBench - Error Group - Unverifiable Premise - Proactive Detection: leaderboard
Metric: Detection Rate (%). Source: github.com. Saturation forecast: Around December 2026. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.5 | 92.55 |
| 2 | DeepSeek V4 Pro | 91.08 |
| 3 | Gemini 3.1 Pro (Preview) | 89.61 |
| 4 | Qwen 3.5 122B A10B | 87.35 |
| 5 | DeepSeek V4 Flash | 84.71 |
| 6 | Qwen 3.5 35B A3B | 83.82 |
| 7 | Qwen 3.5 397B A17B | 82.55 |
| 8 | Qwen 3.5 Plus | 82.16 |
| 9 | Claude Sonnet 4.6 | 74.22 |
| 10 | Llama 3.1 70B Instruct | 41.57 |
| 11 | Llama 3.1 8B Instruct | 35.69 |
Interactive version: theaggregate.ai/benchmark?slug=rpcbench-error-group-unverifiable-premise-proactive-detection · How It Works · Data refreshed daily, snapshot 2026-09-24.