Baseten announced a partnership with OpenAI to offer open-model inference through the OpenAI B2B Marketplace.

Baseten has become one of the first open-model inference providers in OpenAI's B2B Marketplace. Through the partnership, OpenAI enterprise customers can use existing OpenAI commitments to access open models served by Baseten, either within Codex or through the Responses API.

The integration targets enterprises running multi-model agentic coding workflows, where engineering teams route tasks to different models based on cost, quality, and performance trade-offs. Agentic coding has emerged as the first multi-model workflow at global scale, with teams using static or adaptive routers to match each job with the optimal model.

Top engineering teams now operate at Pareto frontiers across intelligence, cost, latency, and capacity. No single model remains the best tool for every task for long. Open and closed models ship on cycles measured in weeks, requiring companies to maintain instant access to the latest options with fast, reliable inference.

Baseten's infrastructure is built for this environment. The company operates more than 90 clusters across over 20 clouds with active-active deployments that reroute traffic instead of stopping work when a provider or region fails. Relationships with over 200 compute providers position Baseten to secure capacity ahead of demand.

Code generation workloads stress every component of the inference stack. Multiple turns, long prompts across large repositories, and massive context windows create performance bottlenecks from tooling down to the engine level. Baseten's full-stack optimization addresses these constraints throughout the pipeline.

The offering includes day-zero model access. When a new open model leads coding benchmarks, Baseten launches it on Model APIs the day it ships, making it as simple as updating a single line of code to evaluate the newest frontier model.

For agentic coding specifically, Blaxel—now part of Baseten—provides every agent its own sandbox to run the code it writes. Enterprise governance features include US-based infrastructure with zero data retention for all prompts, region-pinned deployments, and fine-grained authentication and authorization controls.

The partnership reflects a broader shift in the AI infrastructure market. As enterprises move from single-model experiments to production multi-model systems, companies controlling the routing layer capture value. Baseten positions itself as the infrastructure layer that enables organizations to optimize across open and closed models without rebuilding their tooling.

For OpenAI enterprise customers, the partnership provides a path to evaluate open models alongside GPT models without separate procurement or integration work. Enterprises can join Baseten's waitlist to access open models natively in Codex through an OpenAI commitment, or contact the company for direct engineering consultations on other use cases.