# Export models with Unsloth Studio

Use [Unsloth Studio](https://unsloth.ai/docs/new/studio) to export, save, or convert models to GGUF, Safetensors, or LoRA for deployment, sharing, or local inference in Unsloth, llama.cpp, Ollama, vLLM, and more. Export a trained checkpoint or convert any existing model.

<div data-with-frame="true"><figure><img src="https://3215535692-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FxhOjnexMCB3dmuQFQ2Zq%2Fuploads%2FrrFY8YczW3dDpfYi1k9f%2FScreenshot%202026-03-15%20at%209.28.19%E2%80%AFPM.png?alt=media&#x26;token=d2729e16-799f-48f0-8b07-0248b93fa599" alt="" width="563"><figcaption></figcaption></figure></div>

{% stepper %}
{% step %}

### Select Training Run

Start by selecting the training run you want to export from. Each run represents a complete training session and may contain multiple checkpoints.

After choosing a run, select the checkpoint to export. A checkpoint is a saved version of the model created during training.

<div data-with-frame="true"><figure><img src="https://3215535692-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FxhOjnexMCB3dmuQFQ2Zq%2Fuploads%2FzB12XFNP3UjoAT1l9vz3%2Fimage.png?alt=media&#x26;token=021b8864-b2c5-4a92-927e-e23350610036" alt="" width="563"><figcaption></figcaption></figure></div>
{% endstep %}

{% step %}

### Select Checkpoint

Later checkpoints typically represent the final trained model, but you can export any checkpoint depending on your needs.

<div data-with-frame="true"><figure><img src="https://3215535692-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FxhOjnexMCB3dmuQFQ2Zq%2Fuploads%2F8VfRPUcY3w6zYfNmAIDn%2Fimage.png?alt=media&#x26;token=42565a7d-e62f-4cf0-bd33-90422f1b2194" alt="" width="560"><figcaption></figcaption></figure></div>
{% endstep %}

{% step %}

### Export Methods

Depending on your workflow, you can export a merged model, LoRA adapter weights, or a GGUF model for local inference.

<div data-with-frame="true"><figure><img src="https://3215535692-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FxhOjnexMCB3dmuQFQ2Zq%2Fuploads%2Fh4sPts9rJhHiGqf0UxIs%2Fimage.png?alt=media&#x26;token=4f1d6a76-bd40-4471-ab8d-0b2fe33d0410" alt=""><figcaption></figcaption></figure></div>

Each export method produces a different version of the model depending on how you plan to run or share it. The table below explains what each option exports.

| Export Type      | Description                                                                                       |
| ---------------- | ------------------------------------------------------------------------------------------------- |
| Merged Model     | **16-bit model** with the LoRA adapter merged into the base weights.                              |
| LoRA Only        | Exports **only the adapter weights**. Requires the original base model.                           |
| GGUF / llama.cpp | Converts the model to **GGUF format** for Unsloth / llama.cpp **/** Ollama / LM Studio inference. |
| {% endstep %}    |                                                                                                   |

{% step %}

### Export / Save Locally

When exporting a model, you can choose where the resulting files should be saved. Models can be downloaded directly to your machine or pushed to the Hugging Face Hub for hosting and sharing.

Save the exported model files directly to your machine. This option is useful for running the model locally, distributing files manually, or integrating with local inference tools.

<div data-with-frame="true"><figure><img src="https://3215535692-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FxhOjnexMCB3dmuQFQ2Zq%2Fuploads%2FfsBaE8V2o69jSyCVGIz4%2Fimage.png?alt=media&#x26;token=4ef3fa06-d25b-424a-91e3-42debd3b6908" alt="" width="325"><figcaption></figcaption></figure></div>
{% endstep %}

{% step %}

### Push to Hub

Upload the exported model to the Hugging Face Hub. This allows you to host, share, and deploy the model from a central repository.

You will need a Hugging Face write token to publish the model.

<div data-with-frame="true"><figure><img src="https://3215535692-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FxhOjnexMCB3dmuQFQ2Zq%2Fuploads%2FrvVnuVUYQWv2nkrgFxpK%2Fimage.png?alt=media&#x26;token=5e0b91fe-5225-4bff-9fa9-ec1fb3867b1a" alt="" width="325"><figcaption></figcaption></figure></div>

{% hint style="success" %}
If you are already authenticated with the Hugging Face CLI, the write token can be left empty.
{% endhint %}
{% endstep %}
{% endstepper %}