Skip to main content

Scaleway Provider Guide

Scaleway is a Tier-2 catalog provider: an OpenAI-compatible endpoint described by one JSON file (src/lib/providers/catalog/scaleway.json) rather than hand-written code. That file is the source of truth for everything on this page.

Verification status: this entry is docs-verified only, not yet live-verified. Scaleway's GET /v1/models needs a key (an unauthenticated GET answers HTTP 401), so the model ids come from the vendor's public supported-models page and the roster has not been checked against the API. No account was created and no API key was used to build this entry. evidence.liveMatrix is null until someone runs the live capability matrix with a real key (see Verification status below).

Key Facts​

  • Provider id: scaleway
  • Protocol: OpenAI-compatible (/chat/completions)
  • Base URL: https://api.scaleway.ai/v1 (the Using Generative APIs page also shows a project-scoped form, https://api.scaleway.ai/<project_id>/v1)
  • Default model: mistral-small-3.2-24b-instruct-2506
  • Models in catalog: 11 — the chat models whose "Available in Serverless?" cell on the supported-models page is Yes and that are not in that page's End of Life table. Left out: models whose cell is No or EOL, the audio transcription and embedding models, and six Serverless-Yes models that the End of Life table lists (qwen3-coder-30b-a3b-instruct, pixtral-12b-2409, gemma-3-27b-it, devstral-2-123b-instruct-2512, holo2-30b-a3b, voxtral-small-24b-2507)
  • Streaming: supported — the Chat Completions reference documents stream with server-sent events and says the last chunk contains data: [DONE]
  • Tool calling: model-dependent — the function-calling page says "All the chat models hosted by Scaleway support function calling.", the supported-models page's attribute tables carry a "Supports function calling" row that reads Yes for each of the eleven catalog ids, and for gpt-oss-120b it says "Currently, this model should be used through the Responses API, because the Chat Completions API does not yet support tool-calling for this model."
  • Tools while streaming: supported — the function-calling page has a "Tool calling with stream mode" section
  • Structured output: supported — the structured-outputs page documents json_schema (schema mode) and json_object (JSON mode) and says "All LLMs in the Scaleway library support Structured outputs and JSON mode."
  • Structured output + tools together: not declared (false) — no combined probe was possible without credentials
  • Embeddings: not declared. Scaleway's compatibility page documents an Embeddings API and the supported-models page lists embedding models; this entry covers chat models
  • Thinking: not declared (false) — the supported-models page lists "Supported reasoning efforts" for eight of the catalog models (see the table below); the provider-level flag stays false
  • Billing: free-with-card (NeuroLink's classification, not a Scaleway statement) — see Billing below
  • Key format: none declared. The vendor's examples call the value SCW_SECRET_KEY

Quick Start​

1. Get an API key​

  1. Visit: https://account.scaleway.com/register?service=console and create a Scaleway account; the quickstart (https://www.scaleway.com/en/docs/generative-apis/quickstart/) lists "A Scaleway account logged into the console" and "A valid API key" as requirements
  2. Create an API key (https://www.scaleway.com/en/docs/iam/credentials/create-api-keys/): click IAM & API keys on the top-right drop-down menu of the Scaleway console, open the API keys tab and click + Generate API key. The page says: "A screen displays showing the access key and secret key for your new API key and reminding you that this is your only chance to securely save the secret key." The quickstart passes the secret key as the OpenAI client's api_key. Querying models requires one of the permission sets GenerativeApisModelAccess, GenerativeApisFullAccess or AllProductsFullAccess (https://www.scaleway.com/en/docs/generative-apis/api-cli/using-generative-apis/)
  3. Billing as the vendor states it: the pricing page (https://www.scaleway.com/en/pricing/model-as-a-service/) says "Try out new models with our free tier: 1 million tokens and 60 minutes of audio transcription."; the FAQ (https://www.scaleway.com/en/docs/generative-apis/faq/) says "After reaching this limit, you will be charged per million tokens processed and per minutes of audio processed."; the create-account page (https://www.scaleway.com/en/docs/account/how-to/create-an-account/) says "Ordering Scaleway resources requires a valid credit card."; the payment-method page (https://www.scaleway.com/en/docs/billing/how-to/add-payment-method/) says "Before you can order Scaleway resources, you must add a payment method to your account."; the rate-limits page (https://www.scaleway.com/en/docs/generative-apis/reference-content/rate-limits/) says "Base limits apply if you registered a valid payment method, and they are increased automatically if you also verify your identity."
  4. Set SCALEWAY_API_KEY in your .env file (the vendor's examples read SCW_SECRET_KEY, which NeuroLink does not)

2. Configure​

export SCALEWAY_API_KEY=your-api-key
export SCALEWAY_MODEL=mistral-small-3.2-24b-instruct-2506   # optional — overrides the default model
export SCALEWAY_BASE_URL=https://api.scaleway.ai/v1   # optional — proxy or gateway; the project-scoped form is not exercised with NeuroLink

3. Use it​

import { NeuroLink } from "@juspay/neurolink";

const neurolink = new NeuroLink();

const result = await neurolink.generate({
  input: { text: "Explain context windows in one paragraph." },
  provider: "scaleway",
  model: "mistral-small-3.2-24b-instruct-2506",
});

console.log(result.content);
# CLI
npx @juspay/neurolink generate "Hello" --provider scaleway

Per-request credentials work as they do for every provider:

await neurolink.generate({
  input: { text: "Hello" },
  provider: "scaleway",
  credentials: { scaleway: { apiKey: process.env.SCALEWAY_API_KEY } },
});

Billing​

Scaleway's pricing page (https://www.scaleway.com/en/pricing/model-as-a-service/) says "Try out new models with our free tier: 1 million tokens and 60 minutes of audio transcription." and, under the price table, "You benefit from a free tier on the first 1,000,000 tokens. You'll be charged from token number 1,000,001." The FAQ (https://www.scaleway.com/en/docs/generative-apis/faq/) says "There is a Free Tier available for Serverless." and lists "1,000,000 tokens for models billed by tokens". Prices on the pricing page are in euros ("Prices before tax."), which is why the catalog carries no pricingPerMTok values.

The create-account page (https://www.scaleway.com/en/docs/account/how-to/create-an-account/) says "Ordering Scaleway resources requires a valid credit card." The payment-method page (https://www.scaleway.com/en/docs/billing/how-to/add-payment-method/) says "Before you can order Scaleway resources, you must add a payment method to your account.", and the rate-limits page says "Base limits apply if you registered a valid payment method, and they are increased automatically if you also verify your identity." The entry records free-with-card, which is NeuroLink's classification of the Free Tier plus those payment-method sentences, not a Scaleway statement.

Models​

ModelContext (as printed)Max output (Serverless)Vision€/M in · outNotes
mistral-small-3.2-24b-instruct-2506 ⭐128k32kyes€0.15 / €0.35NeuroLink default; the id named in the Chat Completions reference's model field description and in the rate-limits page's /chat/completions example.
llama-3.3-70b-instruct100k (Serverless)16kno€0.90 / €0.90Fallback; the id in the Chat Completions reference's example request. The vendor calls it "text-only (text in/text out)".
qwen3-235b-a22b-instruct-2507250k16kno€0.75 / €2.25Fallback.
gemma-4-26b-a4b-it256k32kyes€0.25 / €0.50Fallback. Reasoning efforts: none, low, medium, high.
qwen3.6-35b-a3b256k32kyes€0.25 / €1.50Fallback. Reasoning efforts: none, low, medium, high.
qwen3.8-27b256k32kyes€0.60 (cached €0.12) / €3.30Reasoning efforts: none, low, medium, xhigh.
qwen3.5-397b-a17b250k16kyes€0.60 / €3.60Reasoning efforts: none, low, medium, high.
deepseek-v4-flash-0731256k32kno€0.40 (cached €0.08) / €0.80Footnote: "During the preview stage, the context is limited to 256k". Reasoning efforts: none, low, high, max.
glm-5.2256k16kno€1.80 / €5.50Footnote: "During the preview stage, the context is limited to 256k". Reasoning efforts: none, high, max.
mistral-medium-3.5-128b180k16kyes€1.50 / €7.50Footnote: "During the preview stage, the context is limited to 180k". Reasoning efforts: none, high.
gpt-oss-120b128k32kno€0.15 / €0.60The vendor says the Chat Completions API does not yet support tool-calling for this model. Reasoning efforts: low, medium, high.

Ids, context windows, maximum outputs, modalities and reasoning efforts come from the supported-models page and prices from the pricing page (region Paris), both retrieved 2026-09-29. The eleven ids in the table appear on both pages, and the supported-models page's Available in Serverless? cell is Yes for these eleven ids. The contextWindow and maxOutputTokens values in the JSON are the decimal forms of the labels the vendor prints (256k → 256000). The three preview-marked models carry status: "preview"; the other statuses are production because the schema requires a value, not because Scaleway states it.

Vision: vision: true marks the models whose Modalities entry lists Vision (mistral-small-3.2-24b-instruct-2506, gemma-4-26b-a4b-it, qwen3.6-35b-a3b, qwen3.8-27b, qwen3.5-397b-a17b, mistral-medium-3.5-128b). The supported-models page says "You will use vision models through the /v1/chat/completions endpoint." No image request was sent. models.visionModel is the default model.

models.defaultContextWindow (100,000) and models.defaultMaxOutputTokens (16,000) are placeholders for ids outside the catalog. The vendor does not publish them; they equal the smallest Serverless context and maximum-output figures among the catalog models.

Fallback order when the default is unavailable: llama-3.3-70b-instruct → qwen3-235b-a22b-instruct-2507 → gemma-4-26b-a4b-it → qwen3.6-35b-a3b. The runtime fallback model name the loader derives (fallbacks[1]) is qwen3-235b-a22b-instruct-2507.

Verification status​

Tier-2 onboarding requires evidence before a provider is accepted, and pnpm run verify:provider-onboarding gates it in CI. This is what the catalog currently records for Scaleway — docs-verified only, not live-verified:

ProbeResult
Rosterunauthenticated GET https://api.scaleway.ai/v1/models answered HTTP 401 on 2026-09-29, so the model ids are taken from the supported-models page (retrieved 2026-09-29); the roster has not been checked against the API
Auth-failure shapeunauthenticated GET /v1/models and GET /v1/chat/completions both answered HTTP 401 with the body {"status":401,"error":"UNAUTHORIZED","message":"missing 'Authorization' header"} (2026-09-29); errorRules carry 401 and the 403, 422 and 429 statuses listed on the Understanding common errors page
Billingthe pricing page and FAQ state a Free Tier of 1,000,000 tokens; the create-account, payment-method and rate-limits pages state what the Quick Start quotes; 2026-09-29, no signup performed
Tools / structured outputdocumented on the function-calling and structured-outputs pages; neither was exercised live, and no combined tools+schema request was sent — structuredOutputWithTools stays false
Live capability sweepnot run. evidence.liveMatrix is null. Before treating this provider as production-ready, run npx tsx test/continuous-test-suite-provider-matrix.ts --provider=scaleway with a real key and record the result.

Do not treat this entry as equivalent to a live-verified Tier-2 provider (e.g. FriendliAI, Novita AI) until that live matrix has been run and evidence.liveMatrix is filled in.

Troubleshooting​

SymptomCauseFix
HTTP 401The vendor's Understanding common errors page lists 401 as "Unauthorized: The authorization header is missing."Set SCALEWAY_API_KEY; the vendor's examples use SCW_SECRET_KEY, which NeuroLink does not read
HTTP 403The vendor's page lists 403 as "Forbidden: Your API key does not exist or does not have the necessary permissions to access the requested resource."The Using Generative APIs page lists the permission sets GenerativeApisModelAccess, GenerativeApisFullAccess and AllProductsFullAccess for querying models
HTTP 422The vendor's page lists 422 as "Model Not Found: The model key is present in the request payload, but the corresponding model is not found."The Chat Completions reference says "Refer to our supported models list or /models endpoint for available models."
HTTP 429The vendor's page lists 429 as "Too Many Requests" (requests per minute) and "Too Many Tokens" (tokens per minute)The rate-limits page lists ways to raise limits, starting with "Verify your identity to automatically increase your rate limit"
Tool calls not returned on gpt-oss-120bThe supported-models page says "Currently, this model should be used through the Responses API, because the Chat Completions API does not yet support tool-calling for this model."NeuroLink's catalog providers call /chat/completions
Structured output ignored with tools attachedstructuredOutputWithTools is false on this entry — untested combinationNeuroLink omits response_format automatically whenever tools are present, before sending

See also​