RPCBench - Error Group - Boundary-Violating Premise - Proactive Detection: leaderboard

Metric: Detection Rate (%). Source: github.com. Saturation forecast: Around December 2026. 11 models tracked.

Top models

#ModelScore
1GPT-5.594.27
2Qwen 3.5 Plus88.07
3Qwen 3.5 397B A17B86.47
4Claude Sonnet 4.685.09
5Qwen 3.5 122B A10B80.73
6Gemini 3.1 Pro (Preview)76.61
7Qwen 3.5 35B A3B74.08
8DeepSeek V4 Pro44.72
9DeepSeek V4 Flash42.2
10Llama 3.1 70B Instruct37.39
11Llama 3.1 8B Instruct29.13

Interactive version: theaggregate.ai/benchmark?slug=rpcbench-error-group-boundary-violating-premise-proactive-detection · How It Works · Data refreshed daily, snapshot 2026-09-24.