Instructions to use mlx-community/Phi-4-mini-instruct-8bit with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use mlx-community/Phi-4-mini-instruct-8bit with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="mlx-community/Phi-4-mini-instruct-8bit", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("mlx-community/Phi-4-mini-instruct-8bit", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("mlx-community/Phi-4-mini-instruct-8bit", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

MLX

How to use mlx-community/Phi-4-mini-instruct-8bit with MLX:

# Make sure mlx-lm is installed
# pip install --upgrade mlx-lm

# Generate text with mlx-lm
from mlx_lm import load, generate

model, tokenizer = load("mlx-community/Phi-4-mini-instruct-8bit")

prompt = "Write a story about Einstein"
messages = [{"role": "user", "content": prompt}]
prompt = tokenizer.apply_chat_template(
    messages, add_generation_prompt=True
)

text = generate(model, tokenizer, prompt=prompt, verbose=True)

Notebooks
Google Colab
Kaggle
Local Apps
LM Studio

vLLM

How to use mlx-community/Phi-4-mini-instruct-8bit with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "mlx-community/Phi-4-mini-instruct-8bit"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "mlx-community/Phi-4-mini-instruct-8bit",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/mlx-community/Phi-4-mini-instruct-8bit

SGLang

How to use mlx-community/Phi-4-mini-instruct-8bit with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "mlx-community/Phi-4-mini-instruct-8bit" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "mlx-community/Phi-4-mini-instruct-8bit",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "mlx-community/Phi-4-mini-instruct-8bit" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "mlx-community/Phi-4-mini-instruct-8bit",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Pi new

How to use mlx-community/Phi-4-mini-instruct-8bit with Pi:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "mlx-community/Phi-4-mini-instruct-8bit"

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "mlx-lm": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "mlx-community/Phi-4-mini-instruct-8bit"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use mlx-community/Phi-4-mini-instruct-8bit with Hermes Agent:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "mlx-community/Phi-4-mini-instruct-8bit"

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default mlx-community/Phi-4-mini-instruct-8bit

Run Hermes

hermes

MLX LM

How to use mlx-community/Phi-4-mini-instruct-8bit with MLX LM:

Generate or start a chat session

# Install MLX LM
uv tool install mlx-lm
# Interactive chat REPL
mlx_lm.chat --model "mlx-community/Phi-4-mini-instruct-8bit"

Run an OpenAI-compatible server

# Install MLX LM
uv tool install mlx-lm
# Start the server
mlx_lm.server --model "mlx-community/Phi-4-mini-instruct-8bit"
# Calling the OpenAI-compatible server with curl
curl -X POST "http://localhost:8000/v1/chat/completions" \
   -H "Content-Type: application/json" \
   --data '{
     "model": "mlx-community/Phi-4-mini-instruct-8bit",
     "messages": [
       {"role": "user", "content": "Hello"}
     ]
   }'

Docker Model Runner
How to use mlx-community/Phi-4-mini-instruct-8bit with Docker Model Runner:
```
docker model run hf.co/mlx-community/Phi-4-mini-instruct-8bit
```

Unexpected and Incoherent Responses from mlx-community/Phi-4-mini-instruct-8bit

by TeluguLLM - opened Feb 28, 2025

Discussion

TeluguLLM

Feb 28, 2025

Description:

When using the mlx-community/Phi-4-mini-instruct-8bit model, I encountered issues where it generates unexpected or incoherent responses. The output does not align with the provided prompt, and the model behavior appears inconsistent. Additionally, I observed a possible degradation in generation quality, with unusually low coherence and relevancy.

Steps to reproduce:

Run the following command:

mlx_lm.generate --model mlx-community/Phi-4-mini-instruct-8bit --prompt "What is your name?"

Observe the generated output.

mlx_lm.generate --model mlx-community/Phi-4-mini-instruct-8bit --prompt "What is your name?"
config.json: 100%|████████████████████████████████████████████████████████████████| 3.30k/3.30k [00:00<00:00, 26.2MB/s]
sample_finetune.py: 100%|█████████████████████████████████████████████████████████| 6.17k/6.17k [00:00<00:00, 17.8MB/s]
model.safetensors.index.json: 100%|███████████████████████████████████████████████| 32.6k/32.6k [00:00<00:00, 22.1MB/s]
modeling_phi3.py: 100%|███████████████████████████████████████████████████████████| 54.3k/54.3k [00:00<00:00, 3.84MB/s]
configuration_phi3.py: 100%|██████████████████████████████████████████████████████| 10.9k/10.9k [00:00<00:00, 8.84MB/s]
added_tokens.json: 100%|███████████████████████████████████████████████████████████████| 249/249 [00:00<00:00, 450kB/s]
special_tokens_map.json: 100%|████████████████████████████████████████████████████████| 587/587 [00:00<00:00, 1.52MB/s]
tokenizer_config.json: 100%|██████████████████████████████████████████████████████| 2.96k/2.96k [00:00<00:00, 6.78MB/s]
merges.txt: 100%|█████████████████████████████████████████████████████████████████| 2.42M/2.42M [00:00<00:00, 2.48MB/s]
tokenizer.json: 100%|█████████████████████████████████████████████████████████████| 15.5M/15.5M [00:01<00:00, 12.6MB/s]
vocab.json: 100%|█████████████████████████████████████████████████████████████████| 3.91M/3.91M [00:01<00:00, 2.84MB/s]
model.safetensors: 100%|██████████████████████████████████████████████████████████| 4.08G/4.08G [02:35<00:00, 26.1MB/s]
Fetching 12 files: 100%|███████████████████████████████████████████████████████████████| 12/12 [02:36<00:00, 13.05s/it]
==========tensors: 100%|██████████████████████████████████████████████████████████| 4.08G/4.08G [02:35<00:00, 26.3MB/s]
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!01<00:00, 2.85MB/s]
==========
Prompt: 8 tokens, 15.097 tokens-per-sec
Generation: 100 tokens, 11.463 tokens-per-sec
Peak memory: 4.116 GB

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment