GPT-4 hallucinates a lot less than 3.5. Same with the Claude Models. This is from personal experience. There are also benchmarks (like TruthfulQA) that try to m

GPT-4 hallucinates a lot less than 3.5. Same with the Claude Models. This is from personal experience. There are also benchmarks (like TruthfulQA) that try to measure hallucinations that show the same thing.