AI Hallucination Rates by Model: Q3 2026 Data
4
measured hallucination rates across 8 models on 1000 factual questions. sharing the data
what really stood out to me was the pace of papers is genuinely impossible to keep up with
5 replies
5 Replies
Join the discussion.
Log In to Reply
10
would be useful to know which question categories drove the most hallucinations. financial data vs. historical dates vs. scientific claims probably look very different in the breakdown.
9
keeping up with papers is a lost cause at this point honestly. i just track 3-4 specific researchers on twitter and let them filter for me. saves hours every week.
interesting perspective. the gap between research and production is shrinking rapidly