Baseten partners with OpenAI to serve open models in Codex and the Responses API

Baseten has joined OpenAI's B2B Marketplace as one of the first open-model inference providers, with a native integration inside Codex. Enterprise customers can now use existing OpenAI commitments for open models served by Baseten, both in Codex and through the Responses API.

What changes for enterprise customers

In its announcement, Baseten says the partnership lets organizations bring open frontier models into workflows they already run on OpenAI. Codex and GPT models are among the most popular choices for scaling code generation, and routing part of that work to open models is meant to deliver more intelligence per dollar against a commitment the customer has already made.

A multi-model workflow at global scale

According to Baseten, agentic coding is currently the most widely adopted AI use case, and leading teams already mix models along the Pareto frontiers of intelligence, cost, latency and capacity. Tasks are funneled through static or adaptive routers to whichever model fits the job best. With the gap between new open and closed releases down to weeks, no single model stays the best tool for every job for long.

Infrastructure, privacy and support

Baseten says it runs on more than 90 clusters across 20-plus clouds, with deployments active-active, so losing a provider or a region reroutes traffic instead of halting work. Relationships with more than 200 compute providers, it adds, keep capacity growing ahead of demand. Major open models launch on its Model APIs the day they ship, so evaluating a new frontier release means changing one line of code.

Baseten says inference runs on US-based infrastructure with zero data retention (ZDR) for all prompts; deployments can be pinned to specific regions, with fine-grained authentication and authorization plus usage tracking across models, users and keys. It also notes that Blaxel, now part of Baseten, gives every agent its own sandbox, and that LangChain trained custom models for LangSmith Engine using Baseten Loops.

OpenAI customers can join the Marketplace interest list to get access to open models natively in Codex.