RPCBench - Error Group - Inconsistent Premise - Fact Fabrication: leaderboard

Metric: Fact Fabrication Rate (%). Source: github.com. Saturation forecast: Estimated already saturated. 11 models tracked.

Top models

#ModelScore
1Qwen 3.5 397B A17B0.7
2Qwen 3.5 Plus0.7
3Qwen 3.5 122B A10B0.86
4Qwen 3.5 35B A3B1.29
5GPT-5.52.84
6Gemini 3.1 Pro (Preview)3.43
7DeepSeek V4 Pro3.86
8DeepSeek V4 Flash4.82
9Claude Sonnet 4.65.14
10Llama 3.1 70B Instruct46.44
11Llama 3.1 8B Instruct55.01

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