Is staik compatible with the OpenAI SDK?

Yes, staik offers an OpenAI-compatible REST API. Just change base_url to https://api.staik.se/v1 in your existing code.

What models are available?

qwen3.6:35b-a3b (262K context), qwen3.5:9b (262K context), gemma4:31b (262K context, vision), bge-m3:latest (embeddings), and kb-whisper-large (Swedish transcription). All run on dedicated GPU hardware in Sweden.

Where is my data stored?

All data is processed on servers in Sweden. No prompts or responses leave the country. Fully GDPR and Schrems II compliant.

API Documentation — staik LLM API

staik offers an OpenAI-compatible REST API on dedicated GPU hardware in Sweden. Your data never leaves the country. Just change base_url in your existing code — everything else works.

Get started in 5 minutes

1. Get an API key

Sign up at /register — you get 100,000 welcome tokens (valid for 30 days) and a key with the sk-st- prefix. Need more? Pick a plan on the pricing page or buy token packs from the dashboard.

2. Make your first request

python

from openai import OpenAI

client = OpenAI(
    api_key="sk-st-your-key",
    base_url="https://api.staik.se/v1",
)

response = client.chat.completions.create(
    model="qwen3.6:35b-a3b",
    messages=[{"role": "user", "content": "Explain what a vector database is."}],
)
print(response.choices[0].message.content)

3. Pick the right model

Model	Vendor	Parameters	Context window	Vision
qwen3.6:35b-a3b	Alibaba	35B MoE	262,144	—
qwen3.5:9b	Alibaba	9.7B	262,144	—
gemma4:31b	Google DeepMind	31B	262,144	✓
bge-m3:latest	BAAI	568M	8,192	embedding
kb-whisper-large	Kungliga biblioteket	1.5B	—	audio (sv)

qwen3.5:35b-a3b is accepted as a backwards-compatible alias for qwen3.6:35b-a3b.

Read more about the models and how to list them programmatically under Models.

Next steps

Chat completions — parameters, system prompt, and vision
Streaming — token-by-token responses via SSE
Claude Code — run Claude Code against staik with two environment variables
Tool calling — let the model call your functions
Plans and quotas — what's included and how the limits work

Data residency

All inference runs on our own GPU servers in Sweden. Prompts and responses never leave the country. See the privacy policy.