Welcome to the PyTorch Foundation monthly newsletter!
Since joining as Executive Director of the PyTorch Foundation earlier this year, I’ve seen clearly that the AI ecosystem is dealing with a massive challenge: scaling diverse hardware architectures while evolving complex software layers across thousands of different organizations simultaneously. This requires hardware and software to co-evolve together, and that’s a global coordination problem only open source can solve.
Projects like PyTorch, vLLM, and Ray are essential across the whole AI model lifecycle. In this community, you’re building the tools developers use to train, optimize, serve, distribute, and run models across heterogeneous hardware. Our job at the PyTorch Foundation is to help you come together to do this work. I’ve been amazed to see how fast this community is growing, and if you’re new here, welcome! Thanks for joining us on this journey!
Cheers,
Mark Collier, Executive Director, PyTorch Foundation
Announcements
Alibaba Cloud Joins the PyTorch Foundation as a Platinum Member 🎉Alibaba Cloud, a global leader in full-stack artificial intelligence services and the team behind the Qwen family of AI models, has joined PyTorch Foundation as a Platinum Member. Drawing on its experience running PyTorch at scale across heterogeneous hardware, Alibaba Cloud will contribute expertise in AI compiler optimization, multi-chip compatibility, and large-scale stability practices to the upstream community. Read the full announcement here 👉 Alibaba join as a Platinum Member
Join the PyTorch Foundation Ambassador Program
PyTorch Foundation is on the lookout for more PyTorch enthusiasts to join the program! We especially welcome applications from contributors in Africa, Latin America, the Middle East, Oceania, Southeast Asia, and Eastern Europe. Nominate yourself or someone else by June 18th here.
Upcoming Events
ExecuTorch Hackathon, San Francisco, June 27-28
Build and optimize real-time AI applications that run directly on Snapdragon-powered mobile devices using ExecuTorch. Participants will build on Samsung Galaxy S25 Ultra devices powered by Snapdragon and learn from Qualcomm and Meta experts through workshops, mentorship, and hands-on support. Apply by June 15th to attend.
PLDI 2026 AI Summit, Boulder, CO, June 15-19
Workshop on “Writing Performance-Portable Kernels Simplified with Helion” provides compiler researchers, kernel authors, and ML systems engineers the opportunity to dive deep into the technology. This is an entirely interactive session where attendees get to write, autotune, and run real Helion kernels live.
PyTorch Conference China 2026, Shanghai, September 8-9
Schedule goes live Wednesday, June 17 – view it here and register here.
PyTorch Conference North America 2026, San Jose, October 20-21
Make sure to register for your early bird ticket before July 31st!
Recent Events
PyTorch 2.12 Release Live Q&A
PyTorch 2.12 introduced major updates across compilation, distributed systems, export, graph capture, and accelerator support. Andrey Talman, Alban Desmaison, Joe Spisak, and moderator Chris Gottbrath joined a live discussion covering the release and answering questions from the community. 🔗 Watch the recording
PyTorch Docathon 2026 Results in 150+ Merged Pull Requests
The Docathon ran from May 5th through May 19th, bringing together more than 260+ registrants and 30+ active participants. In this blog, we highlight the top contributors.
PyTorch Conference Europe 2026
From PyTorch and vLLM to DeepSpeed, Ray, Helion, and Safetensors, PyTorch Conference North America brings together the open source AI community to share technical advances and collaborate across the AI stack. Hear directly from attendees about their experience in Prague. 👉 What PyTorch Conference Europe 2026 Was Really Like
MLSys 2026
PyTorch Foundation had a booth at MLSys 2026 and saw great foot traffic throughout. One key message stood out at the event: The next phase of AI progress is systems-driven. Models matter, but so do the systems that train, serve, optimize, verify, deploy, and operate them efficiently.
In the News
- Alibaba Cloud joins PyTorch Foundation as a Platinum Member, ITBrief US
- Can LinkedIn’s PyTorch-Powered DuaLip Redefine Web-Scale Optimization? Futurum
PyTorch Case Study
How LinkedIn Uses PyTorch to Solve Extreme-Scale Optimization Problems
In our latest case study, LinkedIn shares how it rebuilt its DuaLip linear programming solver in PyTorch, achieving order-of-magnitude speedups and efficient multi-GPU scaling for optimization workloads.
Latest Blogs
vLLM and PyTorch Work Together to Improve the Developer Experience on aarch64
With PyTorch 2.11, it is possible to install CUDA-enabled PyTorch wheels on aarch64 Linux directly from PyPi. In this post, Kaichao You explains how this improves the installation experience for vLLM users.
TLX Block Attention: A Warp-Specialized Blackwell Kernel for Fixed-Block Sparse Self-Attention
In this post, Meta presents the design of TLX Block Attention, a Triton kernel targeting NVIDIA Blackwell GPUs.
Up to 580tps! New Speed Record of Qwen3.5-397B-A17B on GPU for Agentic Workloads with TokenSpeed
The TokenSpeed inference engine achieved a record-breaking 580 tps running the Qwen3.5-397B-A17B model on GPUs. In this blog post, the TokenSpeed team provides a technical breakdown of how this was achieved.
Using Muon Optimizer with DeepSpeed
In this post, the DeepSpeed team shares a deep dive into the integration setup of Muon Optimizer, implementation of hybrid optimizer strategies, and early benchmark results.
Subscribe to the PyTorch Foundation Newsletter
Get updates directly to your inbox: https://pytorch.org/newsletter/