Artificial Analysis Omniscience Accuracy
General QA · 2025-11-16
Artificial Analysis Omniscience Accuracy measures how often a model answers fact-checking or knowledge questions correctly in the Omniscience evaluation. It is paired with a hallucination-rate metric.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Fable 5 | 61.4 |
| GPT-5.6 Sol | 58.5 |
| GPT-5.5 | 56.9 |
| Gemini 3 Pro Preview | 55.9 |
| Gemini 3.1 Pro Preview | 55.3 |
| Claude Opus 5 | 54.2 |
| Gemini 3 Flash Preview | 54.0 |
| Grok 4.5 | 52.0 |
| Gemini 3.5 Flash | 51.9 |
| GPT-5.3-Codex | 51.8 |
| Grok Build 0.1 | 51.4 |
| Gemini 3.6 Flash | 50.2 |
| GPT-5.4 | 50.0 |
| Opus 4.8 | 46.6 |
| Opus 4.6 | 46.4 |
| Kimi K3 | 46.0 |
| GPT-5.6 Terra | 45.9 |
| Opus 4.7 | 45.8 |
| Opus 4.5 | 45.7 |
| GPT-5.5 Instant | 45.5 |
| Muse Spark | 44.6 |
| GPT-5.5 Instant (2026-06-25 hosted snapshot) | 44.1 |
| GPT-5.2 | 43.8 |
| DeepSeek-V4-Pro | 43.3 |
| GPT-5.6 Luna | 41.5 |
| Grok 4 | 41.4 |
| GPT-5.2-Codex | 40.7 |
| GPT-5 | 40.6 |
| Muse Spark 1.1 | 40.6 |
| Sonnet 4.6 | 40.0 |
| Inkling | 40.0 |
| GPT-5.1-Codex | 39.2 |
| Gemini 2.5 Pro | 39.0 |
| GPT-5-Codex | 38.7 |
| Kimi K2.7 Code | 38.6 |
| o3 | 38.4 |
| DeepSeek-V3.2-Speciale | 38.4 |
| Sonnet 5 | 38.3 |
| Qwen3.6-Max-Preview | 37.7 |
| GPT-5.1 | 37.6 |