SambaNova

What it is for

SambaNova Cloud is a hosted Playground desk: you pick a model, send a prompt, and pay the printed $/1M row. Use it when the job is a SambaCloud chat you start at a public $, not when you need a $20 chat seat.

What makes it unique: the cover is SambaNova’s official Playground (Llama-4-Maverick-17B-128E-Instruct, river-valley image analysis, Image Details with calm water / granite peaks / blue sky) from the community still on Blazing Fast Image Analysis Using Native Multimodal Llama 4 — not a pricing card. The reply itself is ops of Playground.

Key features

• Playground — a signed-in composer plus a model reply (cover still; cloud.sambanova.ai/playground). One desk. Cover Llama-4-Maverick-17B-128E-Instruct / river-valley Image Details sit here.
• Agents — agent runs next to Playground (docs; cover sidebar). An agent, not a chat bubble.
• AI Starter Kits — example apps you clone (docs; github.com/sambanova/ai-starter-kit). A kit, not a Playground reply.
• Inference API — OpenAI-shaped /v1 chat completions (quickstart; OpenAI compatibility). An HTTP call, not the composer.
• Supported Models — the SambaCloud catalog with Production / Preview rows (models; pricing). A catalog, not a chat turn.

View Code / Clear on Playground are CTAs (not desks). Free / Developer / Enterprise tiles are the meter. Cached / input / output $/1M rows are the extra meter, not a desk we fold into Playground. API keys and the SambaNova SDK are install, not a sixth desk. Llama / DeepSeek / gpt-oss swaps are the same Playground pipeline.

Who it's for

Best fit: developers who will read a printed $/1M row (gpt-oss-120b $0.22 / $0.59) and open a first Playground, then stay on Developer pay-as-you-go — not Enterprise unless they need production rate limits / BYOC.

Pricing

Verified 2026-10-23 (Asia/Shanghai) from cloud.sambanova.ai/pricing and cloud.sambanova.ai/plans. Token figures below are the printed USD per 1M tokens (cached input / input / output). The cover’s Llama 4 Maverick row is a snapshot — today’s table did not print it.

Plans:

• Free — start exploring SambaNova Cloud. Add a payment method and purchase credits to run requests. Pay-as-you-go credits. Production models. Community support. No printed welcome-credit $ on the page we fetched — do not invent one.
• Developer — pay as you go. Standard rate limits. Production and Preview models. Community support. Add a card to upgrade.
• Enterprise — contact sales. Production rate limits. Standard support. Subscription pricing. On-request models / BYOC / custom rate limits as add-ons.

Serverless examples on the fetched pricing table (price per 1M tokens):

• MiniMax-M3.1-Flash — cached $0.03 / input $0.30 / output $1.20
• DeepSeek-V3.1 — input $3.00 / output $4.50
• DeepSeek-V3.2 — input $3.00 / output $4.50
• gemma-4-31B-it — input $0.38 / output $1.15
• gpt-oss-120b — input $0.22 / output $0.59
• Meta-Llama-3.3-70B-Instruct — input $0.60 / output $1.20

Confirm checkout; taxes follow your location. Copy the live row you will call.

Limitations / costs

• Free does not print a $0 inference allowance. You add a card and buy credits before the first billed request.
• Rows move. Cover Maverick is gone from today’s public table; pin a current Model ID.
• A Playground desk, not Groq’s Supported Models table. For that price table see Groq.

Alternatives

• Groq — a Supported Models table when you want Groq’s GPT-OSS $0.15 / $0.60, not SambaCloud Playground.
• Cerebras — Instant AI Chat when you want Cerebras’ public token rows, not MiniMax $0.30 / $1.20.
• Together AI — a hosted catalog when you want Together’s printed $/1M cards, not SambaNova Developer pay-as-you-go.

Official tutorial

Official SambaNova channel video: Deep Research Agent Demo: Fast AI with SambaCloud (oEmbed author: SambaNova). First-party SambaCloud agent film; the Playground still is the cover. The agent on the video is a second desk, not generate-on-Playground. We also searched @SambaNovaSystems / @SambaNovaAI — recent uploads are hardware / CEO interviews; this is the official SambaCloud product film we oEmbedded. @sambanova did not resolve.

Written path: cloud.sambanova.ai/pricing plus the Playground still.

Website

https://cloud.sambanova.ai

Repository

https://github.com/sambanova

Source

Catalog row curated for AI Tools Nav. Cover is SambaNova’s official Playground still (Llama-4-Maverick-17B-128E-Instruct + river-valley Image Details). Dollars checked on cloud.sambanova.ai/pricing (MiniMax-M3.1-Flash $0.30 / $1.20; gpt-oss-120b $0.22 / $0.59; Llama-3.3-70B $0.60 / $1.20). Tutorial oEmbed author: SambaNova.

Last verified

2026-10-23 (Asia/Shanghai) editorial check.