Baseten launches Loops SDK for fine-tuning GLM-5.3

baseten · x · 2026-08-29

Baseten has released "Loops", a training SDK that now supports post-training and fine-tuning for GLM-5.3 and GLM-5.3-Flash. Loops allows users to write Python training loops that call dedicated trainers and samplers via the API, supporting LoRA fine-tuning or asynchronous RL at long sequence lengths. Checkpoints can be directly deployed to the Baseten Inference Stack.

Original post →

More from Infra

Infra channel →