Gemini 3 Pro leads five of ten index evals and has the best knowledge accuracy, but an 88% hallucination rate and premium pricing make it among the costliest to run.
Key takeaways · AI-distilled
Artificial Analysis' November 2025 Gemini 3 Pro report separated knowledge accuracy from hallucinationWhen a model states something false with full confidence — inventing facts, citations, or APIs that don't exist.Full definition →. The model led AA-Omniscience accuracy but had an 88% hallucination rate, which limited its lead on the combined index.
Gemini 3 Pro used fewer tokenThe chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.Full definition → than Gemini 2.5 Pro in Artificial Analysis' November 2025 evalA repeatable test for AI quality — a set of tasks plus scoring — used the way software teams use test suites, because model output is too variable to judge by eyeballing.Full definition → but cost 12% more to run the index. Higher input and output prices outweighed the token-efficiency improvement.
The November 2025 Gemini 3 Pro analysis reported support for text, images, video and audio. Its cited MMMU-Pro lead demonstrated image-input reasoning; that result alone did not measure audio or video performance.
Terms in this piece · Glossary
eval — A repeatable test for AI quality — a set of tasks plus scoring — used the way software teams use test suites, because model output is too variable to judge by eyeballing.
token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.
hallucination — When a model states something false with full confidence — inventing facts, citations, or APIs that don't exist.
