mlx-community/DeepSeek-V3.2-mlx-5bit

This model mlx-community/DeepSeek-V3.2-mlx-5bit was converted to MLX format from deepseek-ai/DeepSeek-V3.2 using mlx-lm version 0.30.0, plus the updates from mlx-lm pull request #693.

You can find more similar MLX model quants for a single Apple Mac Studio M3 Ultra with 512 GB at https://huggingface.co/bibproj

Should you wish to use the chat mode, e.g. for translations, instead of the reasoning mode, you have to pass a chat template configuration to the model. For the CLI you can do it like so:
--chat-template-config '{"thinking_mode":"chat"}'

---

Use with mlx

pip install mlx-lm

from mlx_lm import load, generate

model, tokenizer = load("mlx-community/DeepSeek-V3.2-mlx-5bit")

prompt = "hello"

if tokenizer.chat_template is not None:
    messages = [{"role": "user", "content": prompt}]
    prompt = tokenizer.apply_chat_template(
        messages, add_generation_prompt=True
    )

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