Baseten, the company that hosts open-weight AI models for developers, says open models running on its infrastructure are heading into OpenAI’s Codex coding agent. The claim comes from a post on Baseten’s own blog, credited to Dannie Herzberg.
The offer lets OpenAI customers reach those models through their existing OpenAI commitment, according to Baseten. The post calls the access native to Codex, and its summary also mentions OpenAI’s Responses API. It is not open yet: interested customers join a waitlist through an OpenAI interest form.
The post names no models, prices, launch date or customers. Baseten’s own statements, not independently checked, cover its capacity (over 90 clusters spread among 20-plus cloud providers) and its privacy promise of zero data retention on prompts.
Baseten argues that top engineering teams already route every coding task to the model that balances price, output quality and speed most favorably. If OpenAI is now willing to seat outside open models next to its own in Codex, that argument has a large customer channel behind it. Until Baseten publishes which models and at what price, buyers cannot judge whether the savings are real.
Announced by Baseten on its own blog. The post carries no publication date.
