Reinforcement fine-tuning research program

Join our expanded Reinforcement fine-tuning (RFT) research program to develop expert AI models for complex, domain-specific tasks.

RFT enhances model reasoning and accuracy in areas like law, finance, healthcare and coding by customizing with high-quality, graded examples — using significantly less data. Apply if you’re a research enterprise in fields such as software engineering, law or finance with objectively verifiable tasks.

Participants gain access to infrastructure and our ML experts to help create your custom model.

You will receive beta RFT API access to test and give feedback before the planned public release in early Q3 2025. Spots are limited, and feedback is required to help improve the API before release.