From the source
Baseten has made Serverless Loops generally available, a training SDK for supervised fine-tuning and reinforcement learning on frontier open models with a direct path to production inference.
- The SDK lets teams train through a Tinker-compatible API, evaluate checkpoints, and deploy the resulting model to Baseten Dedicated Inference, with Baseten handling GPUs, kernels, model architectures and the numerical correctness of the distributed training infrastructure.
- The initial release covers a selection of popular open-weight models for post-training at a 256k context length for both SFT and RL, billed per token so customers pay only for what they use.
- Trained models can be deployed to Baseten's production inference platform with a single click, which loads the adapter alongside its base model and creates an endpoint on the same stack that serves 15B+ inference requests per day, or weights can be downloaded in
Hugging Face Safetensors format.
- For this GA release Baseten says it improved handling of transient network errors and GPU failures, reduced startup overhead and accelerated weight transfers between trainers and samplers, supporting more concurrent workloads.
NVIDIA, Datadog, LangChain Labs and Ambience Healthcare have already used Loops; NVIDIA post-trained its Nemotron Super 3.5 VL model and improved intent classification accuracy from 76.8% to 94%.





