RPCBench - Domain - Goodreads Dual-Domain - Evidence Distortion: leaderboard

Metric: Evidence Distortion Rate (%). Source: github.com. Saturation forecast: Around January 2028. 11 models tracked.

Top models

#ModelScore
1Gemini 3.1 Pro (Preview)16.74
2GPT-5.518
3Claude Sonnet 4.622.55
4Qwen 3.5 397B A17B22.78
5Qwen 3.5 Plus22.89
6Qwen 3.5 122B A10B23.01
7Llama 3.1 8B Instruct24.94
8DeepSeek V4 Pro25.17
9DeepSeek V4 Flash25.17
10Qwen 3.5 35B A3B25.63
11Llama 3.1 70B Instruct25.74

Interactive version: theaggregate.ai/benchmark?slug=rpcbench-domain-goodreads-dual-domain-evidence-distortion · How It Works · Data refreshed daily, snapshot 2026-09-24.