SmolLM3-3B-GGUF/README.md

---
base_model: HuggingFaceTB/SmolLM3-3B
license: apache-2.0
library_name: transformers
model_creator: HuggingFaceTB
model_name: SmolLM3-3B
quantized_by: Second State Inc.
language:
- en
- fr
- es
- it
- pt
- zh
- ar
- ru
---

<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<img src="https://github.com/LlamaEdge/LlamaEdge/raw/dev/assets/logo.svg" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</div>
<hr style="margin-top: 1.0em; margin-bottom: 1.0em;">
<!-- header end -->

# SmolLM3-3B-GGUF

## Original Model

[HuggingFaceTB/SmolLM3-3B](https://huggingface.co/HuggingFaceTB/SmolLM3-3B)

## Run with LlamaEdge

- LlamaEdge version: [v0.24.0](https://github.com/LlamaEdge/LlamaEdge/releases/tag/0.24.0) and above

- Prompt template

  - Normal chat

    - Prompt type: `smol3-no-think`
  
    - Prompt string
  
      ```text
      <|im_start|>system
      ## Metadata

      Knowledge Cutoff Date: June 2025
      Today Date: 15 July 2025
      Reasoning Mode: /no_think
      
      ## Custom Instructions
      
      You are a helpful AI assistant named SmolLM, trained by Hugging Face.
      <|im_end|>
      <|im_start|>user
      {user_message_1}
      <|im_end|>
      <|im_start|>assistant
      {assistant_message_1}
      <|im_end|>
      <|im_start|>user
      {user_message_2}
      <|im_end|>
      <|im_start|>assistant
      ```

  - Chat with Tools
 
    - Prompt type: `smol3-no-think`
  
    - Prompt string
  
      ```text
      <|im_start|>system
      ## Metadata

      Knowledge Cutoff Date: June 2025
      Today Date: 15 July 2025
      Reasoning Mode: /no_think
      
      ## Custom Instructions
      
      You are a helpful AI assistant named SmolLM, trained by Hugging Face.
      
      ### Tools
      
      You may call one or more functions to assist with the user query.
      You are provided with function signatures within <tools></tools> XML tags:
      
      <tools>
      {"name":"sum","description":"Calculate the sum of two numbers","parameters":{"$schema":"http://json-schema.org/draft-07/schema#","properties":{"a":{"description":"The left hand side number","format":"int32","type":"integer"},"b":{"description":"The right hand side number","format":"int32","type":"integer"}},"required":["a","b"],"title":"SumRequest","type":"object"}}
      {"name":"sub","description":"Calculate the difference of two numbers","parameters":{"$schema":"http://json-schema.org/draft-07/schema#","properties":{"a":{"description":"The left hand side number","format":"int32","type":"integer"},"b":{"description":"The right hand side number","format":"int32","type":"integer"}},"required":["a","b"],"title":"SubRequest","type":"object"}}
      {"name":"get_current_weather","description":"Get the weather for a given city","parameters":{"$schema":"http://json-schema.org/draft-07/schema#","definitions":{"TemperatureUnit":{"enum":["celsius","fahrenheit"],"type":"string"}},"properties":{"location":{"description":"the city to get the weather for, e.g., 'Beijing', 'New York', 'Tokyo'","type":"string"},"unit":{"$ref":"#/definitions/TemperatureUnit","description":"the unit to use for the temperature, e.g., 'celsius', 'fahrenheit'"}},"required":["location","unit"],"title":"GetWeatherRequest","type":"object"}}
      </tools>
      
      For each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:
      <tool_call>
      {"name": <function-name>, "arguments": <args-json-object>}
      </tool_call>

      <|im_end|>
      <|im_start|>user
      {user_message_1}
      <|im_end|>
      <|im_start|>assistant
      {assistant_message_1}
      <|im_end|>
      <|im_start|>user
      {user_message_2}
      <|im_end|>
      <|im_start|>assistant
      ```


  <!-- - `think` mode

    - Prompt type: `smol3-think`

    - Prompt string

      ```text
      <|im_start|>system
      /think

      ## Custom Instructions
      {system_message}
      <|im_end|>
      <|im_start|>user
      {user_message_1}
      <|im_end|>
      <|im_start|>assistant
      <think>
      {think content}
      </think>
      {assistant_message_1}
      <|im_end|>
      <|im_start|>user
      {user_message_2}
      <|im_end|>
      <|im_start|>assistant
      <think>
      {think content}
      </think>
      {assistant_message_2}
      <|im_end|> -->
      ```

- Context size: `128000`

- Run as LlamaEdge service

  ```bash
  wasmedge --dir .:. --nn-preload default:GGML:AUTO:SmolLM3-3B-Q5_K_M.gguf \
    llama-api-server.wasm \
    --model-name SmolLM3-3B \
    --prompt-template smol3-no-think \
    --ctx-size 128000
  ```

- Run as LlamaEdge command app

  ```bash
  wasmedge --dir .:. --nn-preload default:GGML:AUTO:SmolLM3-3B-Q5_K_M.gguf \
    llama-chat.wasm \
    --prompt-template smol3-no-think \
    --ctx-size 128000
  ```

## Quantized GGUF Models

| Name | Quant method | Bits | Size | Use case |
| ---- | ---- | ---- | ---- | ----- |
| [SmolLM3-3B-Q2_K.gguf](https://huggingface.co/second-state/SmolLM3-3B-GGUF/blob/main/SmolLM3-3B-Q2_K.gguf)     | Q2_K   | 2 | 1.25 GB| smallest, significant quality loss - not recommended for most purposes |
| [SmolLM3-3B-Q3_K_L.gguf](https://huggingface.co/second-state/SmolLM3-3B-GGUF/blob/main/SmolLM3-3B-Q3_K_L.gguf) | Q3_K_L | 3 | 1.69 GB| small, substantial quality loss |
| [SmolLM3-3B-Q3_K_M.gguf](https://huggingface.co/second-state/SmolLM3-3B-GGUF/blob/main/SmolLM3-3B-Q3_K_M.gguf) | Q3_K_M | 3 | 1.57 GB| very small, high quality loss |
| [SmolLM3-3B-Q3_K_S.gguf](https://huggingface.co/second-state/SmolLM3-3B-GGUF/blob/main/SmolLM3-3B-Q3_K_S.gguf) | Q3_K_S | 3 | 1.43 GB| very small, high quality loss |
| [SmolLM3-3B-Q4_0.gguf](https://huggingface.co/second-state/SmolLM3-3B-GGUF/blob/main/SmolLM3-3B-Q4_0.gguf)     | Q4_0   | 4 | 1.81 GB| legacy; small, very high quality loss - prefer using Q3_K_M |
| [SmolLM3-3B-Q4_K_M.gguf](https://huggingface.co/second-state/SmolLM3-3B-GGUF/blob/main/SmolLM3-3B-Q4_K_M.gguf) | Q4_K_M | 4 | 1.92 GB| medium, balanced quality - recommended |
| [SmolLM3-3B-Q4_K_S.gguf](https://huggingface.co/second-state/SmolLM3-3B-GGUF/blob/main/SmolLM3-3B-Q4_K_S.gguf) | Q4_K_S | 4 | 1.82 GB| small, greater quality loss |
| [SmolLM3-3B-Q5_0.gguf](https://huggingface.co/second-state/SmolLM3-3B-GGUF/blob/main/SmolLM3-3B-Q5_0.gguf)     | Q5_0   | 5 | 2.16 GB| legacy; medium, balanced quality - prefer using Q4_K_M |
| [SmolLM3-3B-Q5_K_M.gguf](https://huggingface.co/second-state/SmolLM3-3B-GGUF/blob/main/SmolLM3-3B-Q5_K_M.gguf) | Q5_K_M | 5 | 2.21 GB| large, very low quality loss - recommended |
| [SmolLM3-3B-Q5_K_S.gguf](https://huggingface.co/second-state/SmolLM3-3B-GGUF/blob/main/SmolLM3-3B-Q5_K_S.gguf) | Q5_K_S | 5 | 2.16 GB| large, low quality loss - recommended |
| [SmolLM3-3B-Q6_K.gguf](https://huggingface.co/second-state/SmolLM3-3B-GGUF/blob/main/SmolLM3-3B-Q6_K.gguf)     | Q6_K   | 6 | 2.53 GB| very large, extremely low quality loss |
| [SmolLM3-3B-Q8_0.gguf](https://huggingface.co/second-state/SmolLM3-3B-GGUF/blob/main/SmolLM3-3B-Q8_0.gguf)     | Q8_0   | 8 | 3.28 GB| very large, extremely low quality loss - not recommended |
| [SmolLM3-3B-f16.gguf](https://huggingface.co/second-state/SmolLM3-3B-GGUF/blob/main/SmolLM3-3B-f16.gguf)       | f16    | 16 | 6.16 GB| |

*Quantized with llama.cpp b5889*