Qwen3 Coder
Hosted on our own DGX Spark in Austria, available through an OpenAI-compatible API.
Try Qwen3 Coder for free
Every new account starts with 10M free tokens, usable for this and every other model. No card, no subscription.
Fact sheet
- Model name in the API
- qwen3-coder:30b
- Size
- 31B dense (Q4_K_M)
- Context
- 256K
- Input
- Text
- Capabilities
- Tools
- License
- Apache 2.0
- Tier
- Code
- Hosting
- loaded on demand
Call
OpenAI-compatible. The alias works as model name too.
{
"model": "qwen3-coder:30b",
"messages": [ … ]
}
What Qwen3 Coder is good for
Writing and refactoring code
Functions, tests and refactorings in common languages, with an explanation of what changed.
Reading large codebases
256K context: whole modules or several files at once, to find bugs or explain how things fit together.
In editors and agents
Usable through the OpenAI-compatible API in tools that allow a custom endpoint, such as Continue, Cline or Aider.
Code that should stay in-house
Runs in Austria on our own hardware. Your source code is not used for training.
What it is not for
- Qwen3 Coder is specialised in code. For general text, research or conversation, a general-purpose model is better.
- It reads text only, no images or screenshots of error messages.
- It is loaded on demand: the first call after a pause takes a few seconds.
Example: Python with the OpenAI library
Swap base URL and key, everything else works as with OpenAI. Streaming works the same way.
from openai import OpenAI
client = OpenAI(base_url="https://allsmartllm.com/v1",
api_key="sk-allsmartllm-…")
answer = client.chat.completions.create(
model="qwen3-coder:30b",
messages=[
{"role": "system", "content": "You are a careful Python developer."},
{"role": "user", "content": "Write a function that validates an IBAN, with tests."},
],
)
print(answer.choices[0].message.content)
Frequently asked questions
Where does Qwen3 Coder run?
On our own DGX Spark machines in Austria. Your code does not leave the EU and is not used for training.
How large is the model?
About 30 billion parameters, of which only a small part is active per step (mixture of experts). That makes it fast for its size.
Can I use it in my editor?
Yes, in any tool that allows an OpenAI-compatible endpoint with a custom URL. Base URL: https://allsmartllm.com/v1.
How much code fits into one request?
The context is 256K tokens, roughly tens of thousands of lines of code.
What licence does the model have?
Apache 2.0. What it generates for you belongs to you.