ALIA 40B Instruct API

Barcelona Supercomputing Center, open modelalia-40b-instructModel source: Hugging Face

Spain's public model, trained in Spanish, Catalan, Basque and Galician.

ALIA 40B Instruct is the large language model trained by the Barcelona Supercomputing Center under Spain's public AI strategy, with Spanish, Catalan, Basque and Galician as first-class languages. GPU Flow serves it unquantised (BF16) on a full B200, through the OpenAI-compatible API, at €0.20 per million input tokens and €2.00 per million output.

Reads in one go

≈ 350 pages

163,840 tokens

Brain size

Mid-sized

40B denso, BF16

Where it runs

Spain (EU)

1× B200

Understands

Text, not images

Chat

What it is good at

Spain's four official languages. Chat, writing and summaries in Spanish, Catalan, Basque and Galician with native quality.
Public sector and regulated use cases. For projects where a European, publicly funded model is a requirement.
Long documents. A context of 163,840 tokens, about 350 pages in one request.

Text only; no tool calling in this version. For code or agents, use Qwen3.8 27B.

Pricing

€ per million tokens, VAT included

You send (input)0.20 €

It answers (output)2.00 €

Re-reading repeats (cache)0.02 €

1,000 chat answers~0.70 €

Analysing a 200 page contract~0.02 €

try it with 5 € freeActivated with your card. Enough for about 7,143 answers.

For your technical team

Compatible with the OpenAI SDK: change the base URL and the key, and that is it.

curl https://api.gpuflow.ai/v1/chat/completions \
  -H "Authorization: Bearer $GPUFLOW_API_KEY" \
  -d '{ "model": "alia-40b-instruct",
        "messages": [{ "role": "user", "content": "Hola" }] }'

Context and limits

Context: 163,840 tokens per request, about 350 pages.
Requests per minute: 1,000 per key, with bursts up to double; raisable to 5,000 from the dashboard. 5,000 in total per account.
Tokens per minute: 2,000,000 input (cached tokens do not count) and 400,000 output, per key.
Prompt per request: Up to 256,000 tokens, raisable on request.
If you go over: A 429 error with a Retry-After header telling you how many seconds to wait.

ALIA 40B Instruct compared with similar models

	ALIA 40B Instruct	Qwen3.8 27B
Best for	Spain's four official languages with native quality	Coding, agents, structured output
Context	≈ 350 pages	≈ 210 pages
Input / output, € per million	0.20 / 2.00	0.10 / 0.30
1,000 chat answers	~0.70 €	~0.14 €

ALIA for native quality in co-official languages and public-sector requirements; Qwen3.8 27B when you also need tools or a lower output price.

Frequently asked questions

Is ALIA 40B Instruct free?

The model weights are open source; running them costs GPU time. On GPU Flow you pay €0.20 per million input tokens, and every new account gets €5 of credit when it adds a card.

What does the ALIA 40B Instruct API cost?

€0.20 per million input tokens, €2.00 per million output and €0.02 per million cached input. VAT included, no monthly fee.

Can I use ALIA 40B Instruct with Claude Code or Cursor?

Yes. Claude Code connects through the Anthropic-compatible endpoint (/v1/messages); Cursor and other tools through the OpenAI-compatible one (/v1/chat/completions).

Where does it run?

On our NVIDIA B200 GPUs, in a data centre in Spain.

Who trained ALIA?

The Barcelona Supercomputing Center (BSC), as part of Spain's public AI strategy. GPU Flow serves the 40B Instruct version through its API.

Does it speak Catalan, Basque and Galician?

Yes. Spanish, Catalan, Basque and Galician are first-class languages in ALIA, not an afterthought.