AI inference on custom SN50 chips. OpenAI-compatible API with fast output speeds. GPT OSS 120B at 600+ tok/s.

At a glance: Sambanova has 5 models listed on Inference Hub. The cheapest text model is GPT OSS 120B at $0.220 input and $0.590 output per 1M tokens, and the most expensive is GLM 5.3 at $2.80 input and $8.80 output per 1M tokens. There is no free tier. Prices are checked against Sambanova's own pricing every week.

Features

OpenAI Compatible
Streaming
Batching
Fine-tuning
Embeddings
Vision
Audio
Function Calling
JSON Mode

Compliance

None listed

Pricing Model

per token

Details

Models 12
API Base https://api.sambanova.ai/v1
LLM Inference

Model Catalog (12)

Model Type Input $/1M Output $/1M Context Speed Status
DeepSeek R1 0528
DeepSeek · 671B MoE
llm $5.00 $7.00 — —
DeepSeek V3
DeepSeek · 671B MoE
llm $3.00 $4.50 — —
DeepSeek V3.1
DeepSeek · 671B MoE
llm $3.00 $4.50 131k —
DeepSeek V3.2
DeepSeek · 671B MoE (37B active)
llm $3.00 $4.50 — —
GLM 5.3
Z.ai
llm $2.80 $8.80 131k —
GPT OSS 120B
OpenAI · 117B MoE (5.1B active)
llm $0.220 $0.590 — —
Gemma 3 12B
Google · 12B
llm $0.200 $0.350 — —
Gemma 4 31B
Google · 27B
llm $0.380 $1.15 262k —
Llama 3.3 70B
Meta · 70B
llm $0.600 $1.20 — —
Llama 4 Maverick
Meta · 400B (17B active)
llm $0.630 $1.80 — —
MiniMax M3
MiniMax
llm $0.600 $2.40 1.0M —
MiniMax M3.1 Flash
MiniMax
llm $0.300 $1.20 1.0M —

Sambanova guides

Sambanova API FAQ

Is the Sambanova API free?

Not as far as we track: Sambanova has no free API tier on record. The cheapest option we track is Gemma 3 12B at $0.20 in / $0.35 out per 1M tokens.

What is the cheapest Sambanova model?

Gemma 3 12B at $0.20 in / $0.35 out per 1M tokens, based on the 12 models we track for Sambanova.

How many models does Sambanova offer?

We track 12 Sambanova models: 12 LLM.

Is the Sambanova API OpenAI-compatible?

Yes. Sambanova exposes an OpenAI-compatible endpoint, so you can use the OpenAI SDKs by changing the base URL and API key.

Run Sambanova? Add a "Prices tracked on Inference Hub" badge to your site, or email hello@inferencehub.org if a price here is wrong.

Subscribe to our newsletter

News and updates, 1–2 emails per month. No spam.