Baseten Partners with OpenAI to Serve Open Models via Codex and Responses API

Baseten has partnered with OpenAI to become one of the first open-model inference providers in the OpenAI B2B Marketplace. This integration allows OpenAI enterprise customers to use their existing OpenAI commitments to access open models served by Baseten, either natively within Codex or via the Responses API.

Multi-Model Agentic Coding

The partnership addresses the shift toward multi-model workflows in agentic coding, where teams route tasks to the best-fit model based on cost, quality, and latency. By integrating Baseten’s infrastructure, organizations can optimize these workflows across both open and closed models. Baseten emphasizes that agentic coding is currently the most widely adopted AI use case, with Codex and GPT models being popular choices for scaling code-generation.

Infrastructure and Capabilities

Baseten provides several key capabilities to support this integration:

• Full-stack performance optimization: Tailored for code generation workloads involving multiple turns, long prompts, and huge context windows.
• Multi-cloud capacity: Runs on more than 90 clusters across 20+ clouds with active-active deployments to ensure reliability and reroute traffic during outages.
• Day zero model access: Major open models are launched on Model APIs the day they ship, allowing immediate evaluation.
• Developer experience: Includes sandboxing for agents via Blaxel, which is now part of Baseten.
• Enterprise governance: Offers US-based infrastructure with zero data retention (ZDR) for all prompts, along with fine-grained authentication and authorization controls.
• Engineering support: Provides forward-deployed engineers for tuning latency and applied researchers for post-training, as demonstrated by LangChain’s use of Baseten Loops.