---
base_model: microsoft/Phi-3.5-MoE-instruct
license: mit
license_link: https://huggingface.co/microsoft/Phi-3.5-MoE-instruct/resolve/main/LICENSE
language:
- multilingual
pipeline_tag: text-generation
library_name: transformers
model_creator: Microsoft
model_name: Phi-3.5-MoE-instruct
quantized_by: Second State Inc.
tags:
- nlp
- code
---

<!-- 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 -->

# Phi-3.5-MoE-instruct-GGUF

## Original Model

[microsoft/Phi-3.5-MoE-instruct](https://huggingface.co/microsoft/Phi-3.5-MoE-instruct)

## Run with LlamaEdge

- LlamaEdge version: [v0.16.0](https://github.com/LlamaEdge/LlamaEdge/releases/tag/0.16.0) and above

- Prompt template

  - Prompt type: `phi-3-chat`

  - Prompt string

    ```text
    <|system|>
    {system_message}<|end|>
    <|user|>
    {user_message_1}<|end|>
    <|assistant|>
    {assistant_message_1}<|end|>
    <|user|>
    {user_message_2}<|end|>
    <|assistant|>
    ```

- Context size: `128000`

- Run as LlamaEdge service

  ```bash
  wasmedge --dir .:. --nn-preload default:GGML:AUTO:Phi-3.5-MoE-instruct-Q5_K_M.gguf \
    llama-api-server.wasm \
    --prompt-template phi-3-chat \
    --ctx-size 128000 \
    --model-name phi-3-moe
  ```

- Run as LlamaEdge command app

  ```bash
  wasmedge --dir .:. --nn-preload default:GGML:AUTO:Phi-3.5-MoE-instruct-Q5_K_M.gguf \
    llama-chat.wasm \
    --prompt-template phi-3-chat \
    --ctx-size 128000
  ```

## Quantized GGUF Models

| Name | Quant method | Bits | Size | Use case |
| ---- | ---- | ---- | ---- | ----- |
| [Phi-3.5-MoE-instruct-Q2_K.gguf](https://huggingface.co/second-state/Phi-3.5-MoE-instruct-GGUF/blob/main/Phi-3.5-MoE-instruct-Q2_K.gguf)     | Q2_K   | 2 | 15.3 GB| smallest, significant quality loss - not recommended for most purposes |
| [Phi-3.5-MoE-instruct-Q3_K_L.gguf](https://huggingface.co/second-state/Phi-3.5-MoE-instruct-GGUF/blob/main/Phi-3.5-MoE-instruct-Q3_K_L.gguf) | Q3_K_L | 3 | 21.7 GB| small, substantial quality loss |
| [Phi-3.5-MoE-instruct-Q3_K_M.gguf](https://huggingface.co/second-state/Phi-3.5-MoE-instruct-GGUF/blob/main/Phi-3.5-MoE-instruct-Q3_K_M.gguf) | Q3_K_M | 3 | 20.0 GB| very small, high quality loss |
| [Phi-3.5-MoE-instruct-Q3_K_S.gguf](https://huggingface.co/second-state/Phi-3.5-MoE-instruct-GGUF/blob/main/Phi-3.5-MoE-instruct-Q3_K_S.gguf) | Q3_K_S | 3 | 18.1 GB| very small, high quality loss |
| [Phi-3.5-MoE-instruct-Q4_0.gguf](https://huggingface.co/second-state/Phi-3.5-MoE-instruct-GGUF/blob/main/Phi-3.5-MoE-instruct-Q4_0.gguf)     | Q4_0   | 4 | 23.6 GB| legacy; small, very high quality loss - prefer using Q3_K_M |
| [Phi-3.5-MoE-instruct-Q4_K_M.gguf](https://huggingface.co/second-state/Phi-3.5-MoE-instruct-GGUF/blob/main/Phi-3.5-MoE-instruct-Q4_K_M.gguf) | Q4_K_M | 4 | 25.3 GB| medium, balanced quality - recommended |
| [Phi-3.5-MoE-instruct-Q4_K_S.gguf](https://huggingface.co/second-state/Phi-3.5-MoE-instruct-GGUF/blob/main/Phi-3.5-MoE-instruct-Q4_K_S.gguf) | Q4_K_S | 4 | 23.8 GB| small, greater quality loss |
| [Phi-3.5-MoE-instruct-Q5_0.gguf](https://huggingface.co/second-state/Phi-3.5-MoE-instruct-GGUF/blob/main/Phi-3.5-MoE-instruct-Q5_0.gguf)     | Q5_0   | 5 | 28.8 GB| legacy; medium, balanced quality - prefer using Q4_K_M |
| [Phi-3.5-MoE-instruct-Q5_K_M.gguf](https://huggingface.co/second-state/Phi-3.5-MoE-instruct-GGUF/blob/main/Phi-3.5-MoE-instruct-Q5_K_M.gguf) | Q5_K_M | 5 | 29.7 GB| large, very low quality loss - recommended |
| [Phi-3.5-MoE-instruct-Q5_K_S.gguf](https://huggingface.co/second-state/Phi-3.5-MoE-instruct-GGUF/blob/main/Phi-3.5-MoE-instruct-Q5_K_S.gguf) | Q5_K_S | 5 | 28.8 GB| large, low quality loss - recommended |
| [Phi-3.5-MoE-instruct-Q6_K.gguf](https://huggingface.co/second-state/Phi-3.5-MoE-instruct-GGUF/blob/main/Phi-3.5-MoE-instruct-Q6_K.gguf)     | Q6_K   | 6 | 34.4 GB| very large, extremely low quality loss |
| [Phi-3.5-MoE-instruct-Q8_0.gguf](https://huggingface.co/second-state/Phi-3.5-MoE-instruct-GGUF/blob/main/Phi-3.5-MoE-instruct-Q8_0.gguf)     | Q8_0   | 8 | 44.5 GB| very large, extremely low quality loss - not recommended |
| [Phi-3.5-MoE-instruct-f16-00001-of-00003.gguf](https://huggingface.co/second-state/Phi-3.5-MoE-instruct-GGUF/blob/main/Phi-3.5-MoE-instruct-f16-00001-of-00003.gguf)     | f16   | 16 | 29.8 GB|  |
| [Phi-3.5-MoE-instruct-f16-00002-of-00003.gguf](https://huggingface.co/second-state/Phi-3.5-MoE-instruct-GGUF/blob/main/Phi-3.5-MoE-instruct-f16-00002-of-00003.gguf)     | f16   | 16 | 29.4 GB|  |
| [Phi-3.5-MoE-instruct-f16-00003-of-00003.gguf](https://huggingface.co/second-state/Phi-3.5-MoE-instruct-GGUF/blob/main/Phi-3.5-MoE-instruct-f16-00003-of-00003.gguf)     | f16   | 16 | 24.5 GB|  |

*Quantized with llama.cpp b4462.*