Claude 3.7 Sonnet 4.4% Old Benchmark Not Tested on New Why

https://www.bookmark-suggest.win/lexis-ai-vs-westlaw-ai-deconstructing-the-hallucination-rate-narrative-1

Discontinued Model Testing and Its Impact on AI Hallucination Benchmarks Why Older Version Data Warrants Caution in Hallucination Analysis As of April 2025, it’s surprising how many AI evaluations still rely heavily on discontinued