RPCBench - Error Group - Inconsistent Premise - Proactive Detection: leaderboard
Metric: Detection Rate (%). Source: github.com. Saturation forecast: Around December 2026. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | DeepSeek V4 Pro | 75.95 |
| 2 | Claude Sonnet 4.6 | 74.99 |
| 3 | Qwen 3.5 397B A17B | 74.02 |
| 4 | Qwen 3.5 122B A10B | 73.75 |
| 5 | Gemini 3.1 Pro (Preview) | 73.59 |
| 6 | Qwen 3.5 Plus | 73.33 |
| 7 | DeepSeek V4 Flash | 69.31 |
| 8 | GPT-5.5 | 69.09 |
| 9 | Qwen 3.5 35B A3B | 68.61 |
| 10 | Llama 3.1 70B Instruct | 21.96 |
| 11 | Llama 3.1 8B Instruct | 17.46 |
Interactive version: theaggregate.ai/benchmark?slug=rpcbench-error-group-inconsistent-premise-proactive-detection · How It Works · Data refreshed daily, snapshot 2026-09-24.