RPCBench - Error Group - Boundary-Violating Premise - Evidence Faithfulness: leaderboard
Metric: Faithfulness Index (%). Source: github.com. Saturation forecast: Estimated already saturated. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 3.5 Plus | 99.54 |
| 2 | Qwen 3.5 397B A17B | 99.2 |
| 3 | GPT-5.5 | 98.39 |
| 4 | Qwen 3.5 122B A10B | 98.17 |
| 5 | Qwen 3.5 35B A3B | 95.53 |
| 6 | Gemini 3.1 Pro (Preview) | 92.43 |
| 7 | Claude Sonnet 4.6 | 91.17 |
| 8 | DeepSeek V4 Pro | 82.22 |
| 9 | DeepSeek V4 Flash | 81.31 |
| 10 | Llama 3.1 70B Instruct | 70.64 |
| 11 | Llama 3.1 8B Instruct | 62.61 |
Interactive version: theaggregate.ai/benchmark?slug=rpcbench-error-group-boundary-violating-premise-evidence-faithfulness · How It Works · Data refreshed daily, snapshot 2026-09-24.