RPCBench - Error Group - Inconsistent Premise - Conditional Localization: leaderboard

Metric: Localization Accuracy (%). Source: github.com. Saturation forecast: Estimated already saturated. 11 models tracked.

Top models

#ModelScore
1Claude Sonnet 4.694.93
2Qwen 3.5 Plus94.16
3Qwen 3.5 397B A17B93.6
4Gemini 3.1 Pro (Preview)93.56
5Qwen 3.5 122B A10B93.36
6DeepSeek V4 Pro92.74
7GPT-5.592.13
8DeepSeek V4 Flash92.04
9Qwen 3.5 35B A3B91.1
10Llama 3.1 70B Instruct65
11Llama 3.1 8B Instruct60.28

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