Self-serve inferenceOpenAI-compatible
Open models in production.
Start with one request.
Test open models for free. Connect your app or coding agent with one OpenAI-compatible endpoint. Try Bielik and Basal.
- Free account and demo, no card
- Drop-in OpenAI SDK replacement
- 1 open chat model, one endpoint
Pick how you want to use it
No account
Try the live demo
Send a real request to Bielik from your browser. No account needed.
Try the demoDevelopers
Integrate the API
One OpenAI-compatible endpoint with streaming and tool calling. Free account, no card.
Make your first requestBusinesses
Run an AI project
Discuss deployment, integration and data requirements with our team.
See enterprise projectsHow it works
From first test to your daily workflow.
- 01Try it freeUse the live demo below. No account or card required.
- 02Get an API keyCreate a free account and an API key in your account panel.
- 03Keep it in an environment variableStore the key in FABRYKA_API_KEY. Never paste it into code or a repository.
- 04Send a requestUse the OpenAI SDK with the Fabryka base URL. Streaming and tool calls work as in OpenAI.
import os
from openai import OpenAI
client = OpenAI(
base_url="https://fabryka.ai/v1",
api_key=os.environ["FABRYKA_API_KEY"],
)
stream = client.chat.completions.create(
model="bielik-11b-v3.1",
messages=[{"role": "user", "content": "Cześć!"}],
stream=True,
)
for chunk in stream:
if chunk.choices:
print(chunk.choices[0].delta.content or "", end="")import OpenAI from "openai";
const client = new OpenAI({
baseURL: "https://fabryka.ai/v1",
apiKey: process.env.FABRYKA_API_KEY,
});
const stream = await client.chat.completions.create({
model: "bielik-11b-v3.1",
messages: [{ role: "user", content: "Cześć!" }],
stream: true,
});
for await (const chunk of stream) {
process.stdout.write(chunk.choices[0]?.delta?.content ?? "");
}curl https://fabryka.ai/v1/chat/completions \
-H "Authorization: Bearer $FABRYKA_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "bielik-11b-v3.1",
"stream": true,
"messages": [{"role": "user", "content": "Cześć!"}]
}'Try it · no account
Test the real API before you sign up.
Send a real request through the public demo. Shared demo limits apply.
The demo stores your request only to run it: the working request is deleted when generation ends, and a stored result on its first retrieval (uncollected results become eligible for deletion 24 hours after completion by default). The chat API's zero data retention does not cover the demo. Do not enter personal or confidential data. Privacy policy
bielik-11b-v3.1ReadyA small question. A real model.
Models
Open models. One endpoint.
The platform serves models developed by Fabryka AI and by other teams. Serving a model does not imply authorship. Browse the model registry at /v1/models.
| Model | Best for | Features | |
|---|---|---|---|
Bielik 11B v3.1bielik-11b-v3.1 | Polish-language tasks | Streaming · tools · full logprobs | |
Basal 2.0 Maxbasal-2.0-max | Typed decisions: classification and routing | Decisions via /v1/systemone |
Models available through the Fabryka API. List checked on November 15, 2026.
Give it to your coding agent
Copy the provider instructions for your coding agent. They are in English, because the agent reads them.
Show the instructions
Configure an OpenAI-compatible provider for this project.
Base URL: https://fabryka.ai/v1
Model: bielik-11b-v3.1
Read the API key from FABRYKA_API_KEY; never print or commit it.
Use /chat/completions with streaming and OpenAI-format tool calls.
Honor Retry-After for HTTP 429, use bounded retries for transient server errors, and stop on 401/402.
Ask before sending private files or tool output. Run one small request and report the observed result.Enterprise projects
AI projects for your organisation.
We deliver custom AI projects for enterprise teams — from a scoped pilot to integration and ongoing development.
Applications and agents
LLM applications, document workflows and agents connected to your data and business processes.
Evaluation and optimisation
Model selection, task-specific evaluation and optimisation for your quality, latency and compute requirements.
Deployment and integration
API integration and deployment designed around your existing systems, infrastructure and data requirements.
Tell us about your process, data and expected outcome.
Scope, deployment requirements and support are agreed per project.
Do I need an account?
For the API, yes: API keys belong to an account, and creating one is free. The demo on this page works without an account.
Do existing API keys still work?
Yes. Existing API keys continue to work.
What about privacy and zero data retention?
Chat completions (/v1/chat/completions, streaming or not) use application-level zero data retention, also with free API keys: prompts and responses are processed in memory, are not stored and are not used to improve models. Account, billing and operational metadata, such as request IDs, model names and token counts, are retained. The demo has separate rules. Data controls
Your next project starts here.
Free API tests · OpenAI-compatible