Based on a transcript excerpt

Sam Witteveen explains that Artificial Analysis's new 5.0 Intelligence Index benchmark shows open models now trailing closed models by 12 points, up from 3. The new tests use 60% private held-out sets and more agentic tasks, so older leaderboards may have flattered local models—viewers are advised to run their own evaluations.

AI-generated: summaries written by AI from the linked sources. How we use AI