The end of fine-tuning: Why evals, context, and traces matter more
OpenAI will wind down its self-serve fine-tuning platform in May 2026, formalizing a split in how AI teams iterate on products. A small fraction of companies continue intensive model training, while most teams now focus …