inferencerlabs/DeepSeek-V4-Pro-MTP-MLX

DeepSeek-V4-Pro MTP

See DeepSeek-V4-Pro with MTP in action: demonstration videos

This draft model contains the extracted Multi-Token Prediction (MTP) layers to be used alongside the DeepSeek-V4-Pro-MLX model as a speculative decoder for improved performance.

Q2.8 Tested on an M3 Ultra 512 GiB RAM and M4 Max using Inferencer app's distributed compute

Distributed inference	~31 tokens/s @ 1000 tokens ~148.5 GiB
Text inference (with MTP)	Untested

Extracted with a modified version of MLX-VLM

For more details see our demonstration videos or visit DeepSeek-V4-Pro.

Disclaimer

We are not the creator, originator, or owner of any model listed. Each model is created and provided by third parties. Models may not always be accurate or contextually appropriate. You are responsible for verifying the information before making important decisions. We are not liable for any damages, losses, or issues arising from its use, including data loss or inaccuracies in AI-generated content.