FaithJudge - FaithBench Summarization: leaderboard
Share of the model's 72 FaithBench article summaries that the FaithJudge judge flags as hallucinated.
Metric: Hallucination rate (%) on 72 FaithBench summaries. Source: raw.githubusercontent.com. Saturation forecast: Estimated already saturated. 35 models tracked. Updated 17 Nov 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | Gemini 2.5 Flash | 19.44 | #467 | |
| 2 | DeepSeek R1 0528 | 20.83 | #431 | |
| 3 | Gemini 2.5 Pro | 25 | #274 | |
| 4 | Gemini 2.5 Pro (Preview 03-25) | 25 | #239 | |
| 5 | Gemini 2.0 Flash (001) | 29.17 | #727 | |
| 6 | GPT-4.1 | 36.11 | #505 | |
| 7 | Qwen 2.5 32B Instruct | 36.11 | #1257 | |
| 8 | O3 (2025-04-16) | Medium | 36.11 | #229 |
| 9 | GPT-4.1 Mini | 37.5 | #793 | |
| 10 | GPT-4.5 (Preview) | 37.5 | #499 | |
| 11 | Claude 3.7 Sonnet (20250219) | 38.89 | #421 | |
| 12 | Jamba 1.6 Mini | 38.89 | #5461 | |
| 13 | Claude Opus 4 (20250514) | Thinking | 38.89 | #300 |
| 14 | glm-4-9B-chat-hf | 38.89 | #3133 | |
| 15 | GPT-4o (2024-11-20) | 40.28 | #864 |
No result here: #1 Claude Fable 5.2, #2 GPT-6.2 Astra, #3 Gemini 4 Carbon.
Interactive version: theaggregate.ai/benchmark?slug=faithjudge-faithbench-summarization · How It Works · Data refreshed daily, snapshot 2026-11-17.