nicolasembleton/Apodex-1.1-mini-MLX-8bit

nicolasembleton/Apodex-1.1-mini-MLX-8bit

This model nicolasembleton/Apodex-1.1-mini-MLX-8bit was converted to MLX format from apodex/Apodex-1.1-mini using mlx-lm version 0.31.3.

Qwen3.5-35B-A3B MoE, 36B params / 3B active.

Use with mlx

pip install mlx-lm
mlx_lm.generate --model nicolasembleton/Apodex-1.1-mini-MLX-8bit --prompt "Hello"

from mlx_lm import load, generate

model, tokenizer = load("nicolasembleton/Apodex-1.1-mini-MLX-8bit")

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)

Architecture

• —Base: Qwen3.5-35B-A3B (MoE)
• —Experts: 256 total, 8 active per token
• —Layers: 40 (hybrid linear/full attention)
• —Context: 262144 max position embeddings
• —Vision tower: Stripped during conversion (text-only)

Known Issues

Metal OOM during long-context prefill on less than 48 GB unified memory. Keep context under 32k tokens.