Soumith Chintala
@soumithchintala
Cofounded and lead @PyTorch at Meta. Also dabble in robotics at NYU. AI is delicious when it is accessible and open-source.
tendon-driven 3D-printed hand from @irmakkguzey and team at the @LerrelPinto lab. * costs $1300 to build, compact human-profile. * the tendons are actually off-the-shelf fish-line, super strong and never break. the plastic parts break before the tendons ever do. * mountable on…
Despite great advances in learning dexterity, hardware remains a major bottleneck. Most dexterous hands are either bulky, weak or expensive. I’m thrilled to present the RUKA Hand — a powerful, accessible research tool for dexterous manipulation that overcomes these limitations!
"Agents... need SO MUCH of your Life Context that you're better off keeping them local and private." Our closing keynote speaker from @aiDotEngineer NYC: an absolute banger from @soumithchintala on why you need open source models like Llama 4 - though he acks that open models…
'water is transparent only within a very narrow band of the electromagnetic spectrum, so living organisms evolved sensitivity to that band, and that's what we now call "visible light". ' (found via HN)
considering Muon is so popular and validated at scale, we've just decided to welcome a PR for it in PyTorch core by default. If anyone wants to take a crack at it... github.com/pytorch/pytorc…
the model keeps on training even when the underlying infra keeps failing....out-of-the-box PyTorch
torchft + TorchTitan: 1200+ failures, no checkpoints, model convergence. A Llama 3 model was trained across 300 L40S GPUs with synthetic failures every 15s. No restarts. No rollbacks. Just asynchronous recovery and continued progress. 📘 hubs.la/Q03t1Z0b0 #PyTorch…
This is a proper Vibe-coding setup for GPU programmers, and can result in getting surprisingly far! I honestly think that if this authoring experience is v1, then v10 might become the normal way GPU experts start writing serious custom kernels! Great work @anneouyang! (finally…
✨ New blog post 👀: We have some very fast AI-generated kernels generated with a simple test-time only search. They are performing close to or in some cases even beating the standard expert-optimized production kernels shipped in PyTorch. (1/6) [🔗 link in final post]
The Hugging Face team recently announced it's going all in on PyTorch 🔥 "We have seen our user base consolidate in PyTorch," says @LysandreJik, Chief Open Source Officer at @huggingface. "Going forward, we're focusing all our efforts on PyTorch to remove a lot of the bloating…
so excited to do this the third time... @runwayml AI Film Festival!
damnnnnnn, that's something.
Say goodbye to the silent era of video generation: Introducing Veo 3 — with native audio generation. 🗣️ Quality is up from Veo 2, and now you can add dialogue between characters, sound effects and background noise. Veo 3 is available now in the @GeminiApp for Google AI Ultra…
The SWE-Bench verified leaderboard has been updated and OpenHands is both number one overall, and the only open source agent in the top 10! swebench.com Read more about our approach of the OpenHands critic here: all-hands.dev/blog/sota-on-s…
1/ Excited to share that I’m taking on the role of leading Fundamental AI Research (FAIR) at Meta. Huge thanks to Joelle for everything. Look forward to working closely again with Yann & team.
the PyTorch Foundation is becoming an umbrella for great AI open-source projects. @vllm_project and @DeepSpeedAI are joining PyTorch as the first two projects! Super excited to be able to bring together multiple like-minded, high-quality projects into the foundation. Props to…
PyTorch Foundation has expanded into an umbrella foundation. @vllm_project and @DeepSpeedAI have been accepted as hosted projects, advancing community-driven AI across the full lifecycle. Supporting quotes provided by the following members: @AMD, @Arm, @AWS, @Google, @Huawei,…
writing out my Keynote at @MLSysConf this year and I'm pretty excited! "Extreme PyTorch: Inside the Most Demanding ML Workloads—and the Open Challenges in Building AI Agents to Democratize Them"
With first Claude and now Gemini playing Pokemon, I was thinking of doing my own game-playing experiment over the weekend. However, I quickly learned that it's very far from the VLA-style "pixels->plan" that I naively thought it was, and wanted to do myself. It's like 90%…
Gemini 2.5 Pro just got the final 8th badge in Pokemon Blue, incredible pace of progress by the world's most powerful model!!! Next up: victory road and final 4 : )
Google's TPUv7 is out! ML accelerator marketing material is usually pretty inscrutable (what numbers are even comparable?), so here I'll explain concretely how this compares with Nvidia. 🧵
Today we're introducing Gen-4, our new series of state-of-the-art AI models for media generation and world consistency. Gen-4 is a significant step forward for fidelity, dynamic motion and controllability in generative media. Gen-4 Image-to-Video is rolling out today to all paid…