castform lets you finetune open-weights models to perform better at your specific tasks. no research background required.
you bring your data and we’ll auto-generate rl environments, build reward signals and run the training loop end-to-end. everything is configurable with our sdk.
why fine-tune?
frontier models are expensive and general. open-weights models are cheaper but need to be adapted to perform well on specific tasks. fine-tuning does that: a model trained on your data, your success criteria, and your output format will outperform a general-purpose model on your task at lower inference cost.
get started
get started right away on our platform.
quickstart use the castform cli and a coding agent to start a training run in minutes sdk overview benchmax, the companion python sdk for full control over your training runs what is rl learn how rl fine-tuning works rag training fine-tune a model for search and retrieval trace-based training train from existing agent logs