# Model Management

{% hint style="info" %}
The **Model Management** page allows users to view and manage a list of supported models for fine-tuning or inference purposes. Models can be downloaded directly from **Hugging Face** repositories. Before using the feature, ensure that your device is connected to the internet, as model downloads require access to external servers.

Additionally, on the first download attempt, users will be prompted to provide an **Access Token** from Hugging Face. This token can be obtained by following Hugging Face's official guidelines for token generation.
{% endhint %}

<figure><img src="/files/Nc5CADYqptsiFdIrv0D9" alt=""><figcaption></figcaption></figure>

## Features

1. **Model List**:
   * Displays all available models, including their names, sizes, and current processing status.
   * Users can search for specific models using the search bar.
2. **Model Details**:
   * **Model Name**: Specifies the name of the model (e.g., `meta-llama/Llama-3.1-8B-Instruct`).
   * **Support**: Use the <mark style="color:blue;">**`Phison`**</mark> tag for full parameters and <mark style="color:purple;">**`Unsloth`**</mark> for the LoRA mechanism. A model tagged with both supports two different LLM fine-tuning approaches.
   * **Size**: Shows the storage size of each model (e.g., `29.93 GiB`, `262.86 GiB`).
   * **Process Status**: Indicates whether the model is ready to use (<mark style="color:green;">**✓**</mark> for ready).
   * **Action**: Allows users to manage the model (e.g., <mark style="color:blue;">**`download`**</mark>, <mark style="color:blue;">**delete**</mark>).
3. **Download Process**:
   * Displays the download progress for models that are currently being fetched.
   * Provides a clear indication of the model size and estimated time remaining.
   * Supports <mark style="color:blue;">**`pause`**</mark>**&#x20;and&#x20;**<mark style="color:blue;">**`resume`**</mark>functionality for interrupted downloads.
4. **Access Token Requirement**:

   * During the initial download, users are required to provide a Hugging Face <mark style="color:blue;">**`Access Token`**</mark> for authentication.

   <figure><img src="/files/K9lFF4oTOaiw2UKMlnQi" alt="" width="563"><figcaption></figcaption></figure>

   * For token generation, visit the Hugging Face website and follow the official instructions to create and retrieve your personal token.

## Supported Models

{% tabs %}
{% tab title="v1.2.x (Coming Soon)" %}

* <mark style="background-color:yellow;">**Full Parameters (Phison Supported)**</mark>

| <mark style="color:blue;">**Model Family**</mark> | <mark style="color:blue;">**Model**</mark> |
| ------------------------------------------------- | ------------------------------------------ |
| Deepseek-ai                                       | DeepSeek-R1-Distill-Llama-70B              |
|                                                   | DeepSeek-R1-Distill-Qwen-32B               |
|                                                   | deepseek-llm-7b-chat                       |
| Google                                            | gemma-3-27b-it                             |
|                                                   | gemma-3-1b-it                              |
|                                                   | gemma-2-9b-it                              |
| Meta-llama                                        | Llama-3.3-70B-Instruct                     |
|                                                   | Llama-3.2-3B-Instruct                      |
|                                                   | Llama-3.1-405B-Instruct                    |
|                                                   | Llama-3.1-8B-Instruct                      |
|                                                   | Llama-3-Taiwan-70B-Instruct                |
| Microsoft                                         | Phi-4-reasoning                            |
|                                                   | phi-4                                      |
|                                                   | Phi-3.5-mini-instruct                      |
| Mistral                                           | Mistral-Small-3.1-24B-Instruct-2503        |
|                                                   | Mixtral-8x22B-Instruct-v0.1                |
|                                                   | Mixtral-8x7B-Instruct-v0.1                 |
| Qwen                                              | Qwen3-32B                                  |
|                                                   | Qwen3-0.6B                                 |
|                                                   | QwQ-32B                                    |
|                                                   | Qwen2.5-72B-Instruct                       |
|                                                   | Qwen2-7B                                   |

* <mark style="background-color:yellow;">**LoRA (Unsloth Supported)**</mark>

<table data-header-hidden><thead><tr><th></th><th></th><th data-hidden></th></tr></thead><tbody><tr><td><mark style="color:blue;"><strong>Model Family</strong></mark></td><td><mark style="color:blue;"><strong>Model</strong></mark></td><td></td></tr><tr><td>Deepseek-ai</td><td>DeepSeek-R1-Distill-Llama-8B</td><td>8B</td></tr><tr><td></td><td>DeepSeek-R1-Distill-Qwen-14B</td><td>14B</td></tr><tr><td></td><td>DeepSeek-R1-Distill-Qwen-1.5B</td><td>1.5B</td></tr><tr><td></td><td>DeepSeek-R1-Distill-Qwen-7B</td><td>7B</td></tr><tr><td>Google</td><td>gemma-3-1b-it</td><td>1B</td></tr><tr><td></td><td>gemma-3-4b-it</td><td>4B</td></tr><tr><td></td><td>gemma-3-12b-it</td><td>12B</td></tr><tr><td>Meta-llama</td><td>Llama-3.1-8B-Instruct</td><td>8B</td></tr><tr><td></td><td>Llama-3.2-1B-Instruct</td><td>1B</td></tr><tr><td></td><td>Llama-3.2-3B-Instruct</td><td>3B</td></tr><tr><td>Mistralai</td><td>Mistral-7B-Instruct-v0.3</td><td>7B</td></tr><tr><td></td><td>Mistral-Nemo-Instruct-2407</td><td>12B</td></tr><tr><td>Qwen</td><td>Qwen3-0.6B</td><td>0.6B</td></tr><tr><td></td><td>Qwen3-0.6B-Base</td><td>0.6B</td></tr><tr><td></td><td>Qwen3-1.7B</td><td>1.7B</td></tr><tr><td></td><td>Qwen3-1.7B-Base</td><td>1.7B</td></tr><tr><td></td><td>Qwen3-4B</td><td>4B</td></tr><tr><td></td><td>Qwen3-4B-Base</td><td>4B</td></tr><tr><td></td><td>Qwen3-8B</td><td>8B</td></tr><tr><td></td><td>Qwen3-8B-Base</td><td>8B</td></tr><tr><td></td><td>Qwen3-14B</td><td>14B</td></tr><tr><td></td><td>Qwen3-14B-Base</td><td>14B</td></tr><tr><td>Microsoft</td><td>Phi-3-mini-4k-instruct</td><td>3.8B</td></tr><tr><td></td><td>Phi-3-medium-4k-instruct</td><td>14B</td></tr><tr><td>IBM-granite</td><td>granite-3.2-2b-instruct</td><td>2B</td></tr><tr><td></td><td>granite-3.2-8b-instruct</td><td>8B</td></tr></tbody></table>
{% endtab %}

{% tab title="v1.1.2" %}

| <mark style="color:blue;">**Model Name**</mark> | <mark style="color:blue;">**Support**</mark> | <mark style="color:blue;">**Size**</mark> |
| ----------------------------------------------- | -------------------------------------------- | ----------------------------------------- |
| google/gemma-2b-it                              | unsloth                                      | 14.03 GiB                                 |
| meta-llama/Llama-3.1-8B-Instruct                | phison, unsloth                              | 29.93 GiB                                 |
| 01-ai/Yi-1.5-34B-Chat                           | phison                                       | 64.06 GiB                                 |
| 01-ai/Yi-1.5-9B-Chat                            | phison                                       | 16.45 GiB                                 |
| HuggingFaceH4/mistral-7b-sft-beta               | unsloth                                      | 26.98 GiB                                 |
| HuggingFaceTB/SmolLM-1.7B-Instruct              | unsloth                                      | 21.10 GiB                                 |
| HuggingFaceTB/SmolLM-135M-Instruct              | unsloth                                      | 1.83 GiB                                  |
| HuggingFaceTB/SmolLM-360M-Instruct              | unsloth                                      | 4.67 GiB                                  |
| MediaTek-Research/Breeze-7B-Instruct-v0\_1      | phison                                       | 27.90 GiB                                 |
| Nexusflow/Starling-LM-7B-beta                   | unsloth                                      | 13.49 GiB                                 |
| NousResearch/Hermes-2-Pro-Mistral-7B            | unsloth                                      | 13.49 GiB                                 |
| NousResearch/Hermes-3-Llama-3.1-8B              | unsloth                                      | 14.97 GiB                                 |
| Qwen/Qwen1.5-0.5B-Chat                          | phison                                       | 1.16 GiB                                  |
| Qwen/Qwen1.5-1.8B-Chat                          | phison                                       | 3.43 GiB                                  |
| Qwen/Qwen1.5-110B-Chat                          | phison                                       | 207.16 GiB                                |
| Qwen/Qwen1.5-14B-Chat                           | phison                                       | 26.40 GiB                                 |
| Qwen/Qwen1.5-4B-Chat                            | phison                                       | 7.37 GiB                                  |
| Qwen/Qwen1.5-72B-Chat                           | phison                                       | 134.66 GiB                                |
| Qwen/Qwen1.5-7B-Chat                            | phison                                       | 14.39 GiB                                 |
| Qwen/Qwen2-0.5B                                 | unsloth                                      | 953.29 MiB                                |
| Qwen/Qwen2-0.5B-Instruct                        | unsloth                                      | 953.29 MiB                                |
| Qwen/Qwen2-1.5B                                 | unsloth                                      | 2.89 GiB                                  |
| Qwen/Qwen2-1.5B-Instruct                        | unsloth                                      | 2.89 GiB                                  |
| Qwen/Qwen2-72B                                  | phison                                       | 135.44 GiB                                |
| Qwen/Qwen2-72B-Instruct                         | phison                                       | 135.44 GiB                                |
| Qwen/Qwen2-7B                                   | phison, unsloth                              | 14.20 GiB                                 |
| Qwen/Qwen2-7B-Instruct                          | unsloth                                      | 14.20 GiB                                 |
| Qwen/Qwen2.5-0.5B-Instruct                      | unsloth                                      | 953.30 MiB                                |
| Qwen/Qwen2.5-1.5B-Instruct                      | unsloth                                      | 2.89 GiB                                  |
| Qwen/Qwen2.5-3B-Instruct                        | unsloth                                      | 5.76 GiB                                  |
| Qwen/Qwen2.5-72B-Instruct                       | phison                                       | 135.44 GiB                                |
| Qwen/Qwen2.5-7B-Instruct                        | unsloth                                      | 14.20 GiB                                 |
| TinyLlama/TinyLlama-1.1B-Chat-v1.0              | unsloth                                      | 2.05 GiB                                  |
| akjindal53244/Llama-3.1-Storm-8B                | unsloth                                      | 14.97 GiB                                 |
| allenai/Llama-3.1-Tulu-3-8B                     | unsloth                                      | 14.97 GiB                                 |
| deepseek-ai/DeepSeek-Coder-V2-Instruct          | phison                                       | 439.11 GiB                                |
| deepseek-ai/DeepSeek-R1-Distill-Llama-8B        | unsloth                                      | 14.97 GiB                                 |
| deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B       | unsloth                                      | 3.32 GiB                                  |
| deepseek-ai/DeepSeek-R1-Distill-Qwen-7B         | unsloth                                      | 14.19 GiB                                 |
| deepseek-ai/deepseek-llm-7b-chat                | phison                                       | 12.88 GiB                                 |
| google/codegemma-7b-it                          | unsloth                                      | 15.92 GiB                                 |
| google/gemma-1.1-2b-it                          | unsloth                                      | 4.69 GiB                                  |
| google/gemma-1.1-7b-it                          | unsloth                                      | 15.92 GiB                                 |
| google/gemma-2-27b-it                           | phison                                       | 50.74 GiB                                 |
| google/gemma-2-9b-it                            | phison                                       | 17.24 GiB                                 |
| google/gemma-7b-it                              | unsloth                                      | 47.74 GiB                                 |
| meta-llama/Llama-2-7b-chat-hf                   | unsloth                                      | 25.10 GiB                                 |
| meta-llama/Llama-3.1-405B-Instruct              | phison                                       | 2.22 TiB                                  |
| meta-llama/Llama-3.1-70B-Instruct               | phison                                       | 262.86 GiB                                |
| meta-llama/Llama-3.2-3B-Instruct                | unsloth                                      | 11.98 GiB                                 |
| meta-llama/LlamaGuard-7b                        | phison                                       | 25.11 GiB                                 |
| meta-llama/Meta-Llama-3-70B-Instruct            | phison                                       | 262.86 GiB                                |
| meta-llama/Meta-Llama-3-8B-Instruct             | phison, unsloth                              | 29.93 GiB                                 |
| mistralai/Mistral-7B-Instruct-v0.1              | unsloth                                      | 27.47 GiB                                 |
| mistralai/Mistral-7B-Instruct-v0.2              | unsloth                                      | 27.47 GiB                                 |
| mistralai/Mistral-7B-Instruct-v0.3              | unsloth                                      | 27.00 GiB                                 |
| mistralai/Mixtral-8x22B-Instruct-v0.1           | phison                                       | 261.95 GiB                                |
| mistralai/Mixtral-8x7B-Instruct-v0.1            | phison                                       | 177.40 GiB                                |
| teknium/OpenHermes-2.5-Mistral-7B               | unsloth                                      | 26.98 GiB                                 |
| yentinglin/Llama-3-Taiwan-70B-Instruct          | phison                                       | 31.43 GiB                                 |
| {% endtab %}                                    |                                              |                                           |

{% tab title="v1.0.0" %}

| <mark style="color:blue;">**Model Name**</mark> | <mark style="color:blue;">**Size**</mark> |
| ----------------------------------------------- | ----------------------------------------- |
| google/gemma-2-9b-it                            | 18.51 GiB                                 |
| meta-llama/Llama-3.1-8B-Instruct                | 29.93 GiB                                 |
| meta-llama/Llama-3.1-70B-Instruct               | 262.91 GiB                                |
| meta-llama/Llama-3.1-405B-Instruct              | 2.22 TiB                                  |
| meta-llama/Llama-2-7b-chat-hf                   | 25.11 GiB                                 |
| meta-llama/Llama-2-13b-chat-hf                  | 48.48 GiB                                 |
| meta-llama/Llama-2-70b-chat-hf                  | 256.96 GiB                                |
| meta-llama/Meta-Llama-3-8B-Instruct             | 29.93 GiB                                 |
| meta-llama/Meta-Llama-3-70B-Instruct            | 262.91 GiB                                |
| yentinglin/Llama-3-Taiwan-70B-Instruct          | 131.45 GiB                                |
| meta-llama/LlamaGuard-7b                        | 25.11 GiB                                 |
| codellama/CodeLlama-7b-Instruct-hf              | 25.11 GiB                                 |
| codellama/CodeLlama-13b-Instruct-hf             | 48.79 GiB                                 |
| codellama/CodeLlama-70b-Instruct-hf             | 257.09 GiB                                |
| mistralai/Mixtral-8x7B-Instruct-v0.1            | 177.39 GiB                                |
| mistralai/Mixtral-8x22B-Instruct-v0.1           | 261.99 GiB                                |
| taide/TAIDE-LX-7B-Chat                          | 12.92 GiB                                 |
| MediaTek-Research/Breeze-7B-Instruct-v1\_0      | 13.95 GiB                                 |
| MediaTek-Research/Breeze-7B-32k-Instruct-v1\_0  | 27.92 GiB                                 |
| tiiuae/falcon-180B-chat                         | 334.41 GiB                                |
| mistralai/Mistral-7B-Instruct-v0.3              | 27.01 GiB                                 |
| Qwen/Qwen2-7B                                   | 14.2 GiB                                  |
| Qwen/Qwen2-7B-Instruct                          | 14.2 GiB                                  |
| Qwen/Qwen2-72B                                  | 135.44 GiB                                |
| Qwen/Qwen2-72B-Instruct                         | 135.44 GiB                                |
| Qwen/Qwen1.5-0.5B-Chat                          | 1.16 GiB                                  |
| Qwen/Qwen1.5-1.8B-Chat                          | 3.43 GiB                                  |
| Qwen/Qwen1.5-4B-Chat                            | 7.37 GiB                                  |
| Qwen/Qwen1.5-7B-Chat                            | 14.39 GiB                                 |
| Qwen/Qwen1.5-14B-Chat                           | 26.41 GiB                                 |
| Qwen/Qwen1.5-72B-Chat                           | 134.68 GiB                                |
| Qwen/Qwen1.5-110B-Chat                          | 207.14 GiB                                |
| deepseek-ai/deepseek-llm-7b-chat                | 12.88 GiB                                 |
| deepseek-ai/deepseek-llm-67b-chat               | 125.56 GiB                                |
| deepseek-ai/deepseek-moe-16b-chat               | 30.5 GiB                                  |
| 01-ai/Yi-1.5-6B-Chat                            | 11.3 GiB                                  |
| 01-ai/Yi-1.5-9B-Chat                            | 16.45 GiB                                 |
| 01-ai/Yi-1.5-34B-Chat                           | 4.07 GiB                                  |
| {% endtab %}                                    |                                           |
| {% endtabs %}                                   |                                           |

## Steps for Access Token

1. Navigate to the Hugging Face website: <https://huggingface.co>.
2. Log in to your Hugging Face account.
3. Go to your account settings and select the **Access Tokens** tab.
4. Generate a new token with the required permissions.
5. Copy the token and provide it when prompted during the model download process in the application.

For more details, refer to the [Hugging Face documentation](https://huggingface.co/docs/hub/security-tokens).

{% hint style="warning" %}
Some models, such as **`LLAMA`**, require user consent for sharing contact information before download.
{% endhint %}


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://genai-studio.gitbook.io/docs/model/model-management.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.