Modular

2025 was a year of breakthroughs at Modular! 🌟 From setting new SOTA benchmarks on NVIDIA Blackwell to empowering enterprise-scale AI with Mammoth, we pushed the boundaries of what’s possible — and none of it would be possible without our incredible community and team. 📖 Explore our journey and celebrate the milestones in our full Year in Review: modular.com/blog/modular-2…

Programming Massively Parallel Processors is the gold standard GPU programming textbook and the 5th edition just dropped. Mojo 🔥 solutions to every exercise are in today's nightly release. Check out the repo and see what GPU programming looks like in Mojo ⬇️ github.com/modular/modula…

The bottleneck for most large open model deployments isn't the model. It's everything around it. Slow first-token latency, unstable tail latency, GPU utilization that falls apart under load, and the operational complexity of keeping it all running. That's the problem we've been solving. DeepSeek V3.1, state-of-the-art throughput, low latency, fully managed on MAX and Mojo 🔥 on NVIDIA Blackwell GPUs. Come get a first look at @NVIDIAGTC, Booth #3004

#NVIDIAGTC is in full swing. We've demoed DeepSeek and FLUX-2 on B200s, talked Mojo kernels with hundreds of developers, and hosted a dinner with the most interesting minds in AI. There's still time left to stop by Booth #3004!

MAX (AKA @iamtimdavis) is at Booth #3004 this week, suited up and ready to talk AI inference, hand you a Modular tote bag, and launch you back into orbit. No rocket science required. @NVIDIAGTC

We hosted an executive dinner at NVIDIA GTC with @Modular. 600+ people signed up, and we had to turn away more than 500. We had a packed house 🔥 This is one of the only dinners where we discuss the impact of AI agents across the entire stack - from infrastructure/systems to context and agentic engineering. Some of the topics we discussed: - Is the future composed of general models or specialized models? - Are agents a systems problem or model problem? - How are non-technical users adopting agent tooling like Claude Cowork? - Where do humans spend the most time on repetitive document work? Massive thank you to the Modular team for partnering with @llama_index on this, and to every single person who showed up. To the 500+ on the waitlist - next time we're getting a bigger venue.

Three great entry points into the same stack, each optimized for a different audience. Get started building on your own, or stop by Booth #3004 at GTC to connect with our team.

The MAX LLM book: build GPT-2 from scratch using our MAX Python API - llm.modular.com

The top three learning resources we're sharing with attendees at our @NVIDIAGTC booth this year: 1. Structured Mojo kernels blog series, 2. Mojo GPU puzzles, and 3. MAX LLM book. Links in thread 🧵 👇

“Everyone should be a GPU programmer.” @clattner_llvm's goal with @Modular: “What Modular is doing is opening up the box. We’re fixing the language problem and the platform problem. "The goal is to let more developers learn modern compute. And to give developers real choice in the hardware they use.” “Those two things unlock the ecosystem.”

Packed house for the @Modular x @llama_index private dinner listening to @jerryjliu0 + @clattner_llvm share their thoughts on everything happening in AI + here at GTC

Follow along for updates: luma.com/gtc-modular

We're live at @NVIDIAGTC! 👋 Find us at Booth #3004, San Jose Convention Center. We'll be here all week running live GPU demos on NVIDIA Blackwell. Come see MAX + Mojo 🔥 in action. #NVIDIAGTC

NVIDIA GTC kicks off today, and our booth opens in 2 hours! Find us at booth #3004 for Mojo kernel demos, image gen with @bfl_ml's FLUX.2 [klein], and more on @nvidia Blackwell all week. What to expect on the @NVIDIAGTC expo floor: modular.com/blog/modular-a…

“LLVM exists so programmers don’t have to understand every detail of the chip.” @clattner_llvm CEO of @Modular and creator of LLVM: “Underneath every computer are chips, and LLVM abstracts over them so software can run across Intel, ARM, and other architectures.” “It started because the old compiler stack couldn’t support things like runtime compilation.” “Over time it grew into the platform that powers modern languages like Rust, Swift, and Julia.” “And today it’s used everywhere from graphics engines to database optimizers.” Abstraction is what unlocks performance.

“Companies don’t want to chase the latest bugs." Chris Lattner, CEO of @Modular, on what AI infrastructure customers actually want: “They want reliable systems so they can focus on their real problems.” “They also want choice: the ability to run workloads across different types of silicon.” “If we can improve total cost of compute by 40%, that’s 40% more capacity you don’t need to buy.”

GPU kernel development doesn't have to mean thousands of lines of interleaved pipeline, memory, and compute logic. Structured Mojo Kernels Part 2 walks through how separating those concerns into three components with hard boundaries simplifies the codebase, makes kernels easier to maintain, and keeps the same structure working across NVIDIA and AMD hardware generations. modular.com/blog/structure…