Mario Neuhold

@mrioa.at

32 Followers 73 Following

Tech, nature and food nerd, retired gamer (CS:S), active esports enthusiast (alpenScene Daedalus), interested in webdev and music. Oh and I'm a Netflix addict.

A diagram showing the AA-Omniscience Hallucination Rate of current models. The diagram description reads:
AA-Omniscience Hallucination Rate (lower is better) measures how often the model answers incorrectly when it should have refused or admitted to not knowing the answer. It is defined as the proportion of incorrect answers out of all non-correct responses, i.e. incorrect / (incorrect + partial answers + not attempted)

The lowest score is 23% being Qwen3.7 Max. Claude models score from 26% (Haiku 4.5), 36% (Opus 4.7 and 4.8) to 46% (Sonnet 4.6). The lowest GPT Model is GPT 5.5, followed by 89% with GPT 5.4. Gemini 3.1 Pro sits in the middle with 50%, Gemini 3.5 Flash with 61%. Another big one is Grok 4.3 which claims the throne amongst propriatary models with only 26%.