SambaNova Cloud (SambaCloud)

Offering overview

Custom silicon underpins SambaNova Cloud: the company's Reconfigurable Dataflow Units (SN40 and the newer SN50 generation) use a three-tier memory design intended to hold many large models resident at once and switch between them quickly. The cloud exposes these through OpenAI-compatible APIs, so existing client code can be pointed at it with a base-URL change. Production models currently listed include MiniMax M2.7, DeepSeek-V3.1, Meta-Llama-3.3-70B-Instruct and gpt-oss-120b, with DeepSeek-V3.2, MiniMax M3 and Gemma 4 31B in preview. Access is tiered: a Developer tier with per-model rate limits, and Managed Subscription and Dedicated tiers arranged through sales. Per-token prices are shown inside the cloud console rather than on a public page, so no list rate is recorded here. The same stack is sold as SambaStack software and SambaRack hardware (SN40-16 and SN50 systems) for customers who want to run inference in their own data centers, with SambaOrchestrator handling autoscaling, load balancing and model management. Typical buyers are developers and enterprises wanting fast open-model inference, plus national labs and sovereign deployments that need on-premises capacity.

Technical specifications

Containers

Unknown

Kubernetes

Unknown

Root access

Unknown

API access

Unknown

Autoscaling

Unknown

Persistent processes

Unknown

AI capabilities

Model serving

DeepSeek-V3.1, DeepSeek-V3.2, Llama 3.3 70B, GPT-OSS 120B, MiniMax M2.7, MiniMax M3, Gemma 4 31B

AI frameworks

Not stated

Inference endpoints

Unknown

Fine-tuning

Unknown

Vector database

Unknown

Model catalog

Unknown

Monitoring

Not stated

Evaluation

Not stated

Security & compliance

Geography

Provider headquarters are separate from offering infrastructure and availability.

Available plans

Pricing can change. Promotional prices are shown separately from standard prices.

No active plans are recorded for this offering.