SambaCloud API

aid: sambanova:sambacloud-api
name: SambaCloud API
description: The SambaCloud API exposes OpenAI-compatible chat completions over SambaNova's RDU-accelerated
  infrastructure. It serves multiple open model families including DeepSeek V3, Llama 3.3 and Llama 4,
  Gemma 3, MiniMax, and gpt-oss, with text and vision capabilities depending on the model. The API is
  consumed via the sambanova-python and sambanova-typescript SDKs and through OpenAI client libraries.
humanURL: https://docs.sambanova.ai
baseURL: https://api.sambanova.ai/v1
tags:
- Inference
- LLM
- Chat Completions
- OpenAI Compatible
- Multimodal
- REST
properties:
- type: Documentation
  url: https://docs.sambanova.ai
- type: GettingStarted
  url: https://docs.sambanova.ai/cloud/docs/get-started
- type: Developer Portal
  url: https://cloud.sambanova.ai
- type: SDK
  url: https://github.com/sambanova/sambanova-python
- type: SDK
  url: https://github.com/sambanova/sambanova-typescript
- type: StarterKits
  url: https://github.com/sambanova/ai-starter-kit
features:
- name: OpenAI-Compatible Endpoints
  description: Chat completions surface compatible with standard OpenAI SDKs for rapid migration of existing
    applications.
- name: High-Throughput RDU Inference
  description: Backed by SN50 RDU silicon optimized for tokens-per-watt on agentic and reasoning workloads.
- name: Open-Weight Model Catalog
  description: Curated catalog covering DeepSeek V3.1/V3.2, Llama 3.3 70B, Llama 4 Maverick, Gemma 3 12B,
    MiniMax M2.7, and gpt-oss 120B.
- name: Vision and Multimodal Models
  description: Llama 4 Maverick and Gemma 3 endpoints support text plus image inputs for multimodal applications.
- name: Custom Checkpoints
  description: SambaStack feature for deploying customer fine-tuned model checkpoints onto RDU silicon.
- name: Sovereign AI Deployment
  description: Regional partner deployments across Australia, Europe, and the UK for data-residency-sensitive
    customers.
- name: AI Starter Kits
  description: Curated example applications and notebooks for RAG, agents, function calling, and document
    understanding.
useCases:
- name: Agentic Inference Workloads
  description: Run long-running, tool-using agent loops on hardware tuned for tokens-per-watt efficiency.
- name: Retrieval-Augmented Generation
  description: Build enterprise RAG pipelines using starter kits and OpenAI client compatibility.
- name: Sovereign and Regulated AI
  description: Deploy in-region or on-prem for finance, government, and regulated enterprise workloads.
- name: Reasoning and Code Generation
  description: Serve DeepSeek and gpt-oss reasoning models at high throughput for coding and research
    assistants.
- name: Vision Document Understanding
  description: Process documents, images, and charts via multimodal Llama and Gemma endpoints.
integrations:
- name: OpenAI SDK
- name: LangChain
- name: LlamaIndex
- name: Hugging Face
- name: Intel
- name: AWS
- name: n8n
- name: Vercel AI SDK
authentication:
- type: API Key
  description: Authorization Bearer token issued from the /apis dashboard on cloud.sambanova.ai.