Glossary · Term

MiniMax

← all terms

Definition

Plain language

A model lab whose open-weight models are used as teachers in distillation pipelines.

As stated in the literature

An AI company whose open-weight foundation models (e.g., MiniMax M2.5) serve as teacher models in agent distillation work like Orchard.

Why it matters: Open-weight teacher models let outside researchers build strong agents without depending on closed APIs.

For example, a research team might distill MiniMax M2.5's agent behavior into a smaller open student model.

Heard on the show

“A team at MiniMax pulled thirty proofs out of a reinforcement-learning run — thirty proofs their automated grader had scored as essentially perfect.”
Episode 133 — How MiniMax Turned a Reward-Hacking Disaster Into Olympiad Gold

Mentioned in 3 episodes

  1. 133
    How MiniMax Turned a Reward-Hacking Disaster Into Olympiad Gold
  2. 092
    When Search Agents Don't Really Search: The Memory Shortcut Hiding in Browsing Benchmarks
  3. 090
    How MiniMax-M2 Bets That Sparsity Plus Verifiable Rewards Can Match Frontier Agents

Related terms