# llmman is now a supported model provider

> Agno now supports llmman for running local models, using the same OpenAI-compatible pattern as its other local-server providers.

- Published: 2026-09-01
- Author: Eric Curtin
- Categories: Changelog
- Canonical: https://www.agno.com/articles/llmman-is-now-a-supported-model-provider
- Markdown: https://www.agno.com/articles/llmman-is-now-a-supported-model-provider.md

We added `llmman` as a model provider for running local models. It follows the same OpenAI-compatible pattern as our other local-server providers, so you point it at a local model and use it like any other.

```python
from agno.agent import Agent
from agno.models.llmman import Llmman

agent = Agent(model=Llmman(id="qwen3:0.6b-q4_K_M"))
# or the string form: Agent(model="llmman:qwen3:0.6b-q4_K_M")
```

llmman mirrors LM Studio, so it accepts a `json_schema` for structured output even though it has no native structured-output mode. The default model is `qwen3:0.6b-q4_K_M`, a 0.6B model around 0.4 GB that runs on a laptop, so you can try the provider without a GPU box.

See [cookbook](https://github.com/agno-agi/agno/tree/main/cookbook/90_models/llmman) for more details.