GLM-5.3 (per 1M tokens)
Input price 2.80 USD per 1M tokens; cached input 0.14 USD per 1M tokens; output 8.80 USD per 1M tokens.
- Cached input: 0.14 USD per 1M tokens
- Output: 8.80 USD per 1M tokens
AI API Tool

PRODUCT OVERVIEW
SambaCloud is SambaNova's full-stack AI inference platform, built on the company's Reconfigurable Dataflow Unit (RDU) chips and Dataflow Architecture to deliver fast, low-latency inference on large open-source models. It supports a range of models including DeepSeek, Llama, Qwen, Gemma, MiniMax, and gpt-oss, with text, image, and audio processing capabilities. Developers can move from other providers such as OpenAI by setting an API key and base URL, then choosing a model, thanks to OpenAI-compatible endpoints. The platform integrates with tools and frameworks including CrewAI, Hugging Face, Cline, and AWS, and is aimed at engineering teams building high-throughput AI applications, agents, and coding assistants. SambaCloud states that it never sees or collects user data or prompts. Access is available through the cloud dashboard and documented quickstart.
Open the SambaCloud dashboard at cloud.sambanova.ai and obtain your SambaNova API key.
Set OPENAI_API_KEY to your SambaNova API key, set the base URL, and choose a supported model to run.
Select from supported open-source models such as DeepSeek, Llama, Qwen, Gemma, MiniMax, or gpt-oss, depending on your modality needs.
Use documented integrations such as CrewAI, Hugging Face, Cline, and AWS to accelerate application development.
Building high-throughput AI applications with low-latency inference
Running agentic workflows and coding assistants on large open-source models
Migrating existing OpenAI-based applications to SambaNova endpoints
Connecting AI frameworks such as CrewAI and Hugging Face to a hosted inference API
SambaCloud is SambaNova's full-stack AI inference platform for running large open-source models using its RDU chips and Dataflow Architecture.
Supported models include DeepSeek, Llama, Qwen, Gemma, MiniMax, and gpt-oss, with text, image, or audio capabilities depending on the model.
Yes. Set OPENAI_API_KEY to your SambaNova API key, set the base URL, and choose a model to run.
Documented integrations include CrewAI, Hugging Face, Cline, and AWS.
Yes, SambaCloud has announced support for the Anthropic Messages API.
Yes, SambaCloud has introduced prompt caching for supported models such as MiniMax M2.7.
SambaCloud states that it never sees or collects any of your data or user prompts.
Documentation is available at docs.sambanova.ai, including a quickstart and supported models list.
SambaCloud prices are usage-based per 1M tokens in USD, listed per model on the official pricing page. The Free plan requires adding a payment method and purchasing credits, so it is not a permanently free tier. The Developer plan is pay-as-you-go with no fixed price. Enterprise pricing is subscription-based via Contact Sales. Prices shown are as of 2026-11-19 and may change.
Input price 2.80 USD per 1M tokens; cached input 0.14 USD per 1M tokens; output 8.80 USD per 1M tokens.
Input price 0.30 USD per 1M tokens; cached input 0.03 USD per 1M tokens; output 1.20 USD per 1M tokens.
Input price 0.38 USD per 1M tokens; output 1.15 USD per 1M tokens.
Input price 0.22 USD per 1M tokens; output 0.59 USD per 1M tokens.
Subscription-based pricing for larger usage; contact sales.
COMMUNITY REVIEWS
No reviews yet.