Qwen3.6-35B-A3B-mlx-nvfp4

An MLX conversion of Qwen/Qwen3.6-35B-A3B quantized to NVFP4 (4-bit FP4, group size 16) for Apple Silicon with mlx-lm.

This is the MLX analog of NVIDIA's nvidia/Qwen3.6-35B-A3B-NVFP4. NVIDIA's checkpoint stores weights in ModelOpt-packed NVFP4 that mlx-lm cannot read directly, so this build was produced by quantizing the bf16 base with MLX's own NVFP4 mode (--q-mode nvfp4 --q-group-size 16).

• Base model: Qwen/Qwen3.6-35B-A3B (Qwen3.5-MoE, Qwen3_5MoeForConditionalGeneration, 35B total / 3B active)
• Format: MLX, NVFP4 (4-bit FP4, group size 16)
• Approx. size on disk: ~19.5 GB
• Converted with: mlx-lm 0.31.2

Note — text-only. The base is multimodal; mlx-lm converts the language model only (vision tower not included). Tokenizer, chat template, and generation_config are included.

Usage

pip install -U mlx-lm
mlx_lm.generate --model pipenetwork/Qwen3.6-35B-A3B-mlx-nvfp4 \
  --prompt "Write a haiku about Apple Silicon." --max-tokens 256

from mlx_lm import load, generate
model, tokenizer = load("pipenetwork/Qwen3.6-35B-A3B-mlx-nvfp4")
messages = [{"role": "user", "content": "Explain mixture-of-experts in one paragraph."}]
prompt = tokenizer.apply_chat_template(messages, add_generation_prompt=True)
print(generate(model, tokenizer, prompt=prompt, max_tokens=512, verbose=True))