Baseten named an NVIDIA adopter
Baseten was listed among the inference providers running NVIDIA’s new Agent Toolkit/OpenShell blueprints at GTC 2026 reported. That positions Baseten to validate NVIDIA inference blueprints in production and ties them directly into the new agent ecosystem launch announced.
Baseten closed a $300 million Series E at a $5 billion valuation on January 23, 2026 (businesswire.com). NVIDIA was an anchor investor in that round and reportedly contributed roughly $150 million to the financing (siliconangle.com). Baseten’s product pages advertise “dedicated inference for high‑scale workloads,” pre‑optimized model APIs, and a model library that includes Nemotron 3 Super and GLM 5 test endpoints. (baseten.co) An NVIDIA case study states Baseten adopted NVIDIA’s Blackwell data‑center GPUs plus the Dynamo inference framework and TensorRT‑LLM on Google Cloud to scale customer deployments. (nvidia.com) NVIDIA’s Agent Toolkit bundles the OpenShell runtime, Nemotron family models and the open AI‑Q blueprint, and the company claims the hybrid frontier+open approach can cut query costs by about 50%. (nvidianews.nvidia.com) NVIDIA’s developer blog describes OpenShell enforcing out‑of‑process policy constraints and sandboxing for long‑running, self‑evolving agents across DGX Spark, DGX Station and RTX machines. (developer.nvidia.com) Baseten’s public case studies list enterprise customers such as Notion and Sourcegraph, showing production inference workloads and enterprise traction that align with deploying agent runtimes at scale. (baseten.co)
