RPCBench - Domain - Goodreads Dual-Domain - Evidence Faithfulness: leaderboard
Metric: Faithfulness Index (%). Source: github.com. Saturation forecast: Around December 2026. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.5 | 87.81 |
| 2 | Qwen 3.5 122B A10B | 86.67 |
| 3 | Qwen 3.5 Plus | 86.5 |
| 4 | Qwen 3.5 397B A17B | 85.42 |
| 5 | Qwen 3.5 35B A3B | 84.91 |
| 6 | Gemini 3.1 Pro (Preview) | 82.63 |
| 7 | DeepSeek V4 Pro | 78.3 |
| 8 | DeepSeek V4 Flash | 77.05 |
| 9 | Claude Sonnet 4.6 | 71.64 |
| 10 | Llama 3.1 70B Instruct | 42.82 |
| 11 | Llama 3.1 8B Instruct | 38.67 |
Interactive version: theaggregate.ai/benchmark?slug=rpcbench-domain-goodreads-dual-domain-evidence-faithfulness · How It Works · Data refreshed daily, snapshot 2026-09-24.