27B model trained for almost nothing
A 27-billion-parameter model achieves strong results with unusually low post-training compute costs.
Also today
Read today's briefing →Get it every morning
Today's top five AI stories, one line each, at 7am Eastern. Free. One-click unsubscribe.