FuseO1-DeepSeekR1-QwQ-SkyT1-Flash-32B-mlx-4Bit

bobig/FuseO1-DeepSeekR1-QwQ-SkyT1-Flash-32B-mlx-4Bit

128k context, Flash for faster answers, more robust than pure QwQ. If you like it thank the maker, sm54.

The Model bobig/FuseO1-DeepSeekR1-QwQ-SkyT1-Flash-32B-mlx-4Bit was converted to MLX format from sm54/FuseO1-DeepSeekR1-QwQ-SkyT1-Flash-32B using mlx-lm version 0.21.5.

Use with mlx

pip install mlx-lm

from mlx_lm import load, generate

model, tokenizer = load("bobig/FuseO1-DeepSeekR1-QwQ-SkyT1-Flash-32B-mlx-4Bit")

prompt="hello"

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

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