GPT-6.2 models hallucinate at 59% and 55%, among the worst on AA-Omniscience
ArtificialAnlys · x · 2026-11-18
Artificial Analysis measures GPT-6.2 Astra and GPT-6.2 Sol hallucinating at 59% and 55% at max effort on AA-Omniscience — among the highest on the board. Net-score leader Claude Fable 5.2 (max) hits 67 with 79% accuracy, while Gemini 4 Hydrogen hallucinates only 8%. Top intelligence clearly comes with a guessing habit.
More from Models
- OpenAI Astra for Law tops Harvey LAB-AA v1.1 legal benchmark at 20.8% — ArtificialAnlys · 2026-11-18
- OpenAI claims Astra for Law cuts hallucinations 20x vs Opus 5.5, 25x vs Fable 5.2 — sherwinwu · 2026-11-18
- OpenAI cuts Ultrafast pricing 50% for 3 months, brings it to Pro 200 subscribers — OpenAIDevs · 2026-11-18
- GPT-6.2 Astra sets Terminal-Bench Science record: +9 points over GPT-6 Astra at under 1/3 cost — OpenAIDevs · 2026-11-18
- OpenAI says start daily coding with GPT-6.2 Sol: near-flagship scores at ~1/5 the cost — OpenAIDevs · 2026-11-18
- GPT-6.2 Astra tops Agents' Last Exam: +5.2 points at under one-third the per-task cost — OpenAIDevs · 2026-11-18