SambaCloud (SambaNova Cloud) is a hosted inference service for large open-weight models, accessed through an API that mirrors the OpenAI Chat Completions format. A request authenticates with a bearer API key created from a dashboard, and existing OpenAI-compatible code can point at the service by changing the base URL and model name.
Models made available through the service include Llama, Qwen, and DeepSeek variants such as DeepSeek-V3.1, and a model can also be loaded directly into a Gradio interface for a no-code demo, including deployment through Hugging Face.
Features
- OpenAI-compatible API: requests use the same shape as the OpenAI Chat Completions API, over curl, Python, or Gradio
- Bearer-key authentication: API keys are created and managed from a dashboard
- Hosted open models: Llama, Qwen, and DeepSeek models, including DeepSeek-V3.1, run on SambaNova's infrastructure
- Gradio deployment: a model can be loaded directly into a Gradio interface for quick demos
