wbkou/Qwen3.5-40B-Claude-4.5-Opus-Distilled-MLX-mxfp8

Qwen3.5-40B-Claude-4.5-Opus-High-Reasoning-Thinking

Quality: quantized (mxfp8, group size: 32, 8.341 bpw)

40 billion parameters (dense, not moe) expanded from Qwen3.5 27B, then trained on Claude 4.6 Opus High Reasoning dataset via Unsloth on local hardware.

96 layers, 1275 Tensors. (50% more than base model of 27B)

Features variable length reasoning ; less complex = shorter, longer for more complex.

Model performance has increased dramatically.

256K context.

More information

---

Source

This model was converted to MLX format from DavidAU/Qwen3.5-40B-Claude-4.5-Opus-High-Reasoning-Thinking using mlx-vlm version 0.4.