OpenAI Discontinues Self-Serve Fine-Tuning Platform, Shifts Model Customization Strategy
Key Takeaways
- ▸OpenAI has discontinued its self-serve fine-tuning platform as of May 7th, 2026
- ▸Inference on existing fine-tuned models remains supported until their base models are deprecated
- ▸The discontinuation is part of OpenAI's broader model retirement and platform consolidation strategy
Summary
On May 7th, 2026, OpenAI announced the discontinuation of its self-serve fine-tuning platform, effective immediately. Developers will no longer be able to create new fine-tuned models through OpenAI's self-service interface. However, inference on previously fine-tuned models will remain available until their corresponding base models are deprecated.
This move is part of OpenAI's broader strategy to retire older models and consolidate its product offerings as the company launches newer, safer, and more capable models. The announcement follows a series of deprecations including legacy GPT model snapshots (April 2026), Sora 2 video generation models (March 2026), and the Assistants API (August 2025). Developers affected by the fine-tuning shutdown have been notified via email with migration guidance and timeline information.
- Developers relying on fine-tuning must plan migration before affected base models reach end-of-life dates
