# PyTorch 2.15 adds FP8 and faster LLM decoding on Apple Silicon MPS

> Original title: MPS in PyTorch 2.15

- Source: Hugging Face Blog
- Published: 2026-10-20T00:00:00.000Z
- Bad Influence: https://badinfluence.news/items/aaqb0i9ot7ufohj279lio4wnw
- Original: https://huggingface.co/blog/mps-pytorch-2-15

## Summary

Hugging Face's PyTorch team says PyTorch 2.15 adds FP8 tensor support on MPS and speeds up mixture-of-experts decoding 5 to 6.5 times. On an M3 Ultra, Qwen3.6-35B-A3B rises from 9.4 to 61.3 tokens/s. The release also cuts peak driver memory for Wan2.2 video VAE decoding from 270 GiB to 8 GiB and requires macOS 15 or later and Python 3.11 or later.

## Why it matters

The post gives measured speedups and memory figures for MPS in PyTorch 2.15, useful for judging whether local Mac inference now suits a given model.