aiXamine - Hallucination: leaderboard
aiXamine Hallucination service score: factuality and faithfulness on SimpleQA, TriviaQA, TruthfulQA, SelfCheckGPT, HaluEval, FaithEval and Vectara tests (aiXamine, arXiv 2608.20554v1, Appendix Table 7). Normalized 0-100 score aggregated from the service's tests; higher is better. 42 models evaluated under identical black-box conditions via OpenRouter or containerized vLLM.
Metric: Service score (0-100; normalized pass rate). Source: arxiv.org. Saturation forecast: Around February 2027. 42 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 Mini | 83.9 | #193 |
| 2 | Gemini 3 Flash (Preview) | 82.7 | #90 |
| 3 | Claude Haiku 4.5 | 80.8 | #287 |
| 4 | GPT-4o | 77.5 | #357 |
| 5 | Gemini 2.0 Flash (001) | 76.2 | #365 |
| 6 | Gemma 4 31B | 75.7 | #212 |
| 7 | DeepSeek R1 0528 | 74.9 | #232 |
| 8 | Claude 3.5 Haiku | 74.9 | #571 |
| 9 | Grok 3 Mini | 74 | #289 |
| 10 | Gemma 4 26B A4B | 73.2 | #360 |
| 11 | Qwen 3.5 122B A10B | 73.1 | #184 |
| 12 | Gemini 2.5 Flash | 72.8 | #255 |
| 13 | Llama 3.3 70B | 72.1 | #578 |
| 14 | Qwen 3.5 27B | 71.5 | #192 |
| 15 | Gemma 4 E2B | 71.2 | #1062 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=aixamine-hallucination · How It Works · Data refreshed daily, snapshot 2026-11-05.