Artificial Analysis releases AA-Omniscience v1.1 with a smaller, recalibrated question set

AIArtificial Analysis released a public set for AA-Omniscience v1.1 on Hugging Face, calibrated to match the difficulty of its private set. At release, public and private Net Scores correlate at 0.99 with a median difference of 0.0 points. Accuracy and hallucination rate also correlate at 0.99, with median differences of -0.07 and -0.29 percentage points.

AIArtificial Analysis releases AA-Omniscience v1.1, a knowledge and hallucination benchmark cut from 6,000 to 3,000 questions, with a newly calibrated 300-question public set. Scores correlate 0.99 between public and private sets, and models average 1.9 points lower on Net Score than on v1. Gemini 4 Carbon (high) leads with a Net Score of 54, ahead of Claude Opus 5.5 (48, max with fallback).

AIArtificial Analysis says AA-Omniscience v1.1 keeps the v1 rank order, with a 0.98 Spearman correlation, and models score about 2 points lower on Net Score on average. The median Net Score drop is 0.6 points, but MiniMax M3.1 Flash falls 22 points, from 16 to -6, and MiMo V2.6 Pro falls 13 points, from 8 to -5.