The whole dayFri 30 Oct 2026
AI Compass

ResearchArtificial AnalysisConfirmed

Story 12 of 12 in the Friday 30 October edition

Artificial Analysis updates AA-Omniscience v1.1 knowledge and hallucination benchmark

Artificial Analysis released AA-Omniscience v1.1 on 30 October 2026, curating 3k factuality questions across 42 topics and publishing a calibrated 300-question public subset on Hugging Face.

Logo: Artificial AnalysisCover by AI Compass

AA-Omniscience v1.1 trims the evaluation from 6k to 3k audited questions across six domains and 42 topics while penalising confident wrong answers over abstentions, accompanied by an open 300-question subset on Hugging Face with 0.99 score correlation. On the updated leaderboard, Gemini 4 Carbon (High) leads Net Score at 54 ahead of Claude Opus 5.5 (48) and Claude Fable 5.1 (46), while Gemini 4 Argon (High) posts the lowest hallucination rate at 4%.

Sources

  1. PrimaryAA-Omniscience Benchmark Leaderboardartificialanalysis.ai · Official methodology and v1.1 model rankings
  2. On XArtificial Analysis (@ArtificialAnlys) on Xx.com · Official v1.1 release thread

The rest of the day

Read the whole day

Spotted an error? Email aiforsuccess.in+compass@gmail.com or tag @ai_for_success on X