Maxime Labonne

Liquid AI · Decision models and Liquid AI's d1

Maxime Labonne, Head of Post-Training at Liquid AI, came back to ThursdAI the week decision models went mainstream. He explained why decision models output no tokens and just pick from a predefined set of answers, walked through Liquid's three d1 releases (an API, a 3B text-and-vision model and a 600M omni model with audio), and was blunt that decision-model benchmarks can't really be trusted yet.

Decision models don't output a single token

Maxime on ThursdAI

“Yes, decision models do not output any tokens. They just take a decision from a predefined set of answers.”

Maxime Labonne · 1:34:44 →

“It's a lot of rebranding, it's true, but it also creates a lot of value”

Maxime Labonne · 1:35:30 →

“But right now, it's, it's really a wild world, and you cannot really trust these benchmarks.”

Maxime Labonne · 1:47:47 →

“I think it's really going to become a primitive. You see it like in SQL queries, you can just like add it like a function calling this model in SQL queries, and it's so cheap that you don't even care, right?”

Maxime Labonne · 1:50:54 →

Suggested posts