RPCBench - Error Group - Inconsistent Premise - Proactive Detection: leaderboard

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

Top models

#ModelScore
1DeepSeek V4 Pro75.95
2Claude Sonnet 4.674.99
3Qwen 3.5 397B A17B74.02
4Qwen 3.5 122B A10B73.75
5Gemini 3.1 Pro (Preview)73.59
6Qwen 3.5 Plus73.33
7DeepSeek V4 Flash69.31
8GPT-5.569.09
9Qwen 3.5 35B A3B68.61
10Llama 3.1 70B Instruct21.96
11Llama 3.1 8B Instruct17.46

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