SambaNova Pricing
Per-token pricing for SambaNova's curated set of 4 frontier open models, served on custom RDU hardware built for high-speed inference.
Last updated: Nov 21, 2026
Get our weekly newsletter on pricing changes, new releases, and tools.
SambaNova offers free models
Free models available — see the full SambaNova free tier guide
Provider | Model | Input/1M | Output/1M |
|---|---|---|---|
$0.220 | $0.590 | ||
MM | $0.300 | $1.200 | |
$0.380 | $1.150 | ||
Z | $2.800 | $8.800 |
Pricing from SambaNova. Prices per 1M tokens.
About SambaNova
SambaNova serves a deliberately small, curated catalog of frontier open models — Llama, DeepSeek, gpt-oss, and similar — on its own custom RDU (Reconfigurable Dataflow Unit) hardware rather than GPUs. Its pitch is speed: very high tokens-per-second on large models, which matters for agents and real-time apps where latency dominates cost.
The API is OpenAI-compatible and includes a free developer tier for testing. Because the catalog is curated rather than exhaustive, SambaNova is best compared on the specific large models it carries — see the per-token prices below and the head-to-head links.
SambaNova Pricing FAQ
How fast is SambaNova?
SambaNova runs models on custom RDU hardware and markets industry-leading tokens-per-second on large open models like Llama 405B and DeepSeek — typically much faster than GPU-based serverless providers.
How much does SambaNova cost?
SambaNova charges per token. Prices vary by model size — see the live table above for current input and output prices on each model it serves.
Does SambaNova have a free tier?
Yes. SambaNova Cloud offers a free developer tier for testing, with paid tiers for higher rate limits and production usage.
Which models does SambaNova offer?
A curated set of frontier open models (Llama, DeepSeek, gpt-oss and similar) rather than an exhaustive catalog. The table above lists the models with current public pricing.
Compare SambaNova with Other Providers
vs
vs
vs
vs
vs
vs
vs
vs
vs
vs
vs
vs