MindLab Open-Sources Macaron-V1, Betting on Continual Learning

MindLab has open-sourced the Macaron-V1 model family, focusing on an Agent-oriented Mixture-of-LoRA approach. Featuring a 748-billion parameter flagship version built on a frozen GLM-5.2, the models are designed around continual learning and multi-LoRA collaboration, representing a major bet on post-training methodologies.

2026-07-23 ~ 2026-07-24 · 4 related posts

• MindLab launches Macaron-V1, a 748B LoRA-MoE agent model aimed at long-horizon tasks — 智东西 · 2026-07-23
• MindLab launches Macaron-V1, a 748B open-weight agent model with 2M-token RL — 青稞AI · 2026-07-24
• MindLab launches Macaron-V1 and says continual learning is the next AI frontier — 新智元 · 2026-07-24
• MindLab open-sources Macaron-V1, a GLM-5.2 family built on continual learning — 机器之心 · 2026-07-24