Back to Tools Directory

MTPLX

Python

3x decode TPS increase On Qwen 3.6 27B @ temp 0.6 | Native MTP Speculative Decoding On Apple Silicon With No External Drafter.

Categories & Topics

anthropic-compatibleapple-siliconinference-enginelocal-aimetalmlxmtpmtplxnative-mtpopenai-compatibleqwenqwen3-nextspeculative-decodingspeculative-sampling

Repository Stats

1.1k
GitHub Stars
84
Forks
1.1k
Watchers
28
Open Issues
License
Apache License 2.0
Repository
youssofal/MTPLX
Last Updated
Today
August 7, 2026
Latest Release
v2.5.4
Today

Related Tools