X · @soumithchintala
· X / Twitter
this looks pretty good for agentic. that it fits on a dgx spark is **chef's kiss**
this looks pretty good for agentic.that it fits on a dgx spark is **chef's kiss**Poolside: Today we're releasing Laguna S 2.1, our most capable model to date.It's a 118B total parameter Mixture-of-Experts model with 8B activated per token, a context window of up to 1M tokens, and thinking and no-thinking modes.Capable