MagicaLabs details its pretraining recipe: tens of multiplicative changes, bug fixes as compute multipliers

magicailabs · x · 2026-09-09

MagicaLabs explains its pretraining recipe is the multiplicative result of tens of changes across model architecture, optimizer, training objective, and data — and that fixing minor bugs proved to be a compute multiplier too. More on the research process in their blog.

Related event: Magic Claims 50x Pretraining Efficiency, Matching DeepSeek V4 Pro at ~$500K(7 posts)→

Original post →

More from Research

Research channel →