Baseten on Hugging Face Inference Providers 🔥

August 6, 2026 | Source: Hugging Face Blog

Tags: Baseten, Hugging Face, inference, open-weight models, DeepSeek, Kimi, GLM

Baseten is now integrated as a supported Inference Provider on the Hugging Face Hub, giving developers serverless access to Kimi K3, DeepSeek V4 Flash, GLM-5.2, and other open-weight models directly from HF model pages and SDKs.

Details

Hugging Face has added Baseten to its Inference Providers ecosystem, enabling developers to run Hub-hosted models through Baseten's infrastructure without leaving the HF interface. Initial support covers conversational and text-generation tasks; additional task types are planned. Developers can authenticate two ways: using their own Baseten API key (calls route directly to Baseten, billed to their account) or routing through HF (charged to the HF account, no separate provider key needed). Provider preference ordering is configurable per user. Open-weight models available at launch include Kimi K3, DeepSeek V4 Flash, and GLM-5.2. The integration also works within popular agent harnesses — Pi, OpenCode, Hermes Agents, and OpenClaw — meaning teams can swap in Baseten-hosted models without additional configuration. Both Python (huggingface_hub >= 1.26.1) and JavaScript (@huggingface/inference) SDKs support the integration.