@MiniMax_AI: hi elie, thanks for your question > 10B active parameters was intentional > M2.5 is getting close t...
hi elie, thanks for your question > 10B active parameters was intentional > M2.5 is getting close to “infinite agent scaling” > knowledge capacity is the main limit > tradeoffs are thoughtfully chosen for efficiency & practicality > pretraining innovations remain exciting areas > structural innovations over just scaling up > Excited for M3 🚀