flâneur

Macaron-V1-Preview: 749B MoL Agent Model post-trained from GLM5.1

macaron.im · 3,080 words · saved by 1 readers

Macaron-V1-Preview

Macaron-V1-Preview: 749B MoL Agent Model post-trained from GLM5.1 We introduce Macaron-V1-Preview , a 749B (744B base + 5 × 1B LoRA) Agent Model post-trained from GLM5.1 with MinT . Macaron-V1-Preview leverages a novel Mixture-of-LoRA (MoL) architecture to provide a scalable, resource-efficient foundation for advanced agentic use cases in general life scenarios. By orthogonalizing the parameter space through the MoL architecture, Macaron-V1-Preview reduces the optimization interference that often appears when one model is trained for diverse agentic capabilities. Each capability is optimized i

saved by

related reading