aiXamine - Overall: leaderboard
Uniform mean of nine aiXamine black-box service scores spanning safety, security and privacy (adversarial and jailbreak robustness, code security, fairness, hallucination, privacy, OOD robustness, over-refusal and safety alignment) for 42 API, open-weight and distilled models (aiXamine, arXiv 2608.20554v1, Appendix Table 7). 0-100; higher is better.
Metric: Mean service score (0-100). Source: arxiv.org. Saturation forecast: Around February 2027. 42 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 Mini | 81.8 | #193 |
| 2 | Gemini 3 Flash (Preview) | 80.6 | #90 |
| 3 | Claude Haiku 4.5 | 80 | #287 |
| 4 | Gemma 4 31B | 80 | #212 |
| 5 | Gemma 4 26B A4B | 78.9 | #360 |
| 6 | GPT-OSS-120B | 78.7 | #346 |
| 7 | GPT-4o | 78 | #357 |
| 8 | DeepSeek R1 0528 | 76.9 | #232 |
| 9 | GPT-OSS-20B | 76.3 | #523 |
| 10 | Qwen 3.5 27B | 76.3 | #192 |
| 11 | Qwen 3.5 122B A10B | 76.1 | #184 |
| 12 | Grok 4 Fast | 76 | #266 |
| 13 | Grok 4.1 Fast | 75.3 | #227 |
| 14 | Phi-4 | 75.1 | #707 |
| 15 | Claude 3.5 Haiku | 75.1 | #571 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=aixamine-overall · How It Works · Data refreshed daily, snapshot 2026-11-05.