Together AI's Vipul Ved Prakash says teams are swapping closed models for smaller ones

Together AI's Vipul Ved Prakash, speaking at TechCrunch Disrupt, says the most capable teams are replacing closed models with smaller, faster ones.

He says the aim is to keep token usage high while cutting costs, which the Together AI post summarizes as "token maxing without cost maxing."

The post does not name any specific models, prices, or benchmark results, so the claim rests on Prakash's statement alone. A video of the talk is linked from the post.