GLM-5.3 MTP

See GLM-5.3 with MTP in action: demonstration videos

This draft model contains the extracted Multi-Token Prediction (MTP) layers from zai-org/GLM-5.3-BF16 for use alongside the GLM-5.3-MLX model as a speculative decoder for improved performance.

Q4-bit quant typically achieves higher throughput with less RAM usage and no loss in quality compared to the base MTP layers.

Tested with Inferencer v2.5.3

M3 Ultra: Q4i without decoder 2,500 tokens~15.25 tokens/s ~427 GiB
M3 Ultra: Q4i with decoder 2,500 tokens~18.1 tokens/s ~431 GiB

Enable speculative decoding in Inference Contols:

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.

Downloads last month
477
MLX
Hardware compatibility
Log In to add your hardware

Quantized

Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for inferencerlabs/GLM-5.3-MTP-MLX

Base model

zai-org/GLM-5.3
Finetuned
(22)
this model