RPCBench - Error Group - Inconsistent Premise - Evidence Distortion: leaderboard
Metric: Evidence Distortion Rate (%). Source: github.com. Saturation forecast: Around December 2026. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.5 | 9.91 |
| 2 | Qwen 3.5 Plus | 11.68 |
| 3 | Qwen 3.5 397B A17B | 12.91 |
| 4 | Qwen 3.5 122B A10B | 12.96 |
| 5 | Gemini 3.1 Pro (Preview) | 14.57 |
| 6 | Qwen 3.5 35B A3B | 16.23 |
| 7 | DeepSeek V4 Pro | 20.35 |
| 8 | DeepSeek V4 Flash | 20.51 |
| 9 | Claude Sonnet 4.6 | 20.73 |
| 10 | Llama 3.1 8B Instruct | 25.66 |
| 11 | Llama 3.1 70B Instruct | 28.92 |
Interactive version: theaggregate.ai/benchmark?slug=rpcbench-error-group-inconsistent-premise-evidence-distortion · How It Works · Data refreshed daily, snapshot 2026-09-24.