--- title: SambaNova Provider Guide description: RDU-accelerated inference — Llama 3.3, GPT-OSS 120B, DeepSeek V3.x, MiniMax, Gemma 4 via SambaNova Cloud keywords: sambanova, rdu, llama-3.3, gpt-oss-120b, deepseek, minimax, gemma, fast inference --- # SambaNova Provider Guide **Open-weight flagships (Llama 3.3 70B, GPT-OSS 120B, DeepSeek V3.x, MiniMax, Gemma 4) on SambaNova's RDU-accelerated cloud** --- ## Overview SambaNova Cloud serves open-weight models on its Reconfigurable Dataflow Units. NeuroLink wraps `api.sambanova.ai/v1` (OpenAI-compatible, zero-quirk Tier 2 catalog entry) so the standard generate / stream contract applies. Roster verified against a live authenticated `/v1/models` on 2026-08-27: - **`Meta-Llama-3.3-70B-Instruct`** (default) — production, 128K context - **`gpt-oss-120b`** — production, 128K context - **`DeepSeek-V3.1`** — production reasoning model, 128K context - **`DeepSeek-V3.2`** — vendor preview, 32K context - **`MiniMax-M2.7`** — production, 192K context - **`MiniMax-M3`** — vision-capable per the vendor dashboard - **`gemma-4-31B-it`** — vendor preview; vision (text + image + video) ### Key Facts - **Protocol**: OpenAI-compatible (`/v1/chat/completions`) - **Default base URL**: `https://api.sambanova.ai/v1` - **Default model**: `Meta-Llama-3.3-70B-Instruct` - **Vision**: `gemma-4-31B-it` (image + video) and `MiniMax-M3` - **Streaming / tool calling**: per the OpenAI-compatible contract - **⚠️ Billing (probed live 2026-08-27)**: new accounts have **no free allowance** — every call returns 402 `PAYMENT_METHOD_REQUIRED` (`balance_units: 0`) until a payment method is added and credits are purchased at [cloud.sambanova.ai/plans/billing](https://cloud.sambanova.ai/plans/billing). The onboarding wizard's payment step may show "temporarily unavailable" — use the Billing page instead. - **Pricing** (per million tokens, vendor pricing page 2026-08-27): Llama-3.3-70B $0.60/$1.20 · gpt-oss-120b $0.22/$0.59 · DeepSeek-V3.x $3.00/$4.50 · MiniMax $0.60/$2.40 · gemma-4-31B-it $0.38/$1.15 - **API keys are bare UUIDs** (no `sk-`-style prefix) — keep them out of logs; NeuroLink's Authorization-header redaction covers them. --- ## Quick Start ### 1. Get an API Key Sign up at [https://cloud.sambanova.ai](https://cloud.sambanova.ai) (Google/Microsoft OAuth), complete the profile step, add a payment method + credits under **Plans → Billing**, and create a key under **API Keys**. ### 2. Configure Environment ```bash # Required SAMBANOVA_API_KEY=... # Optional: override the default model SAMBANOVA_MODEL=gpt-oss-120b # Optional: override the base URL # SAMBANOVA_BASE_URL=https://api.sambanova.ai/v1 ``` ### 3. Generate Your First Response ```typescript import { NeuroLink } from "@juspay/neurolink"; const ai = new NeuroLink(); const result = await ai.generate({ provider: "sambanova", input: { text: "What is a reconfigurable dataflow unit?" }, }); console.log(result.content); ``` --- ## CLI Usage ```bash pnpm run cli generate "Quick question" --provider sambanova pnpm run cli stream "Count to ten" --provider sambanova pnpm run cli generate "Hi" --provider sambanova --model gpt-oss-120b ``` --- ## Configuration Reference | Environment Variable | Required | Default | Description | | -------------------- | -------- | ----------------------------- | ----------------- | | `SAMBANOVA_API_KEY` | Yes | — | SambaNova API key | | `SAMBANOVA_MODEL` | No | `Meta-Llama-3.3-70B-Instruct` | Default model | | `SAMBANOVA_BASE_URL` | No | `https://api.sambanova.ai/v1` | Base URL | --- ## Troubleshooting ### 402 on every call / "PAYMENT_METHOD_REQUIRED" The account has no credits — new accounts have no free allowance. Add a payment method and purchase credits at [cloud.sambanova.ai/plans/billing](https://cloud.sambanova.ai/plans/billing). ### "Invalid SambaNova API key" ```bash test -n "$SAMBANOVA_API_KEY" && echo "SAMBANOVA_API_KEY is set" || echo "SAMBANOVA_API_KEY is missing" ``` Rotate at [cloud.sambanova.ai/apis](https://cloud.sambanova.ai/apis). A bad key returns an OpenAI-shaped 401 with `"code": "invalid_api_key"`. ### 404 for a model that used to exist Preview models (`DeepSeek-V3.2`, `gemma-4-31B-it`) "may be removed at short notice" per the vendor. Verify with an authenticated `GET https://api.sambanova.ai/v1/models`. --- ## See Also - [Cerebras Provider](/docs/getting-started/providers/cerebras) — sibling wafer-scale speed provider - [Groq Provider](/docs/getting-started/providers/groq) — sibling LPU speed provider - [Tier 2 catalog entry guide](/docs/provider-integration/tiers/tier-2-catalog-entry) — internal wiring --- **Need Help?** Open a [GitHub Discussion](https://github.com/juspay/neurolink/discussions) or [issue](https://github.com/juspay/neurolink/issues).