Modular

?? Announcing MAX 24.5 - delivering state of the art Llama 3.1 token generation on CPUs! MAX 24.5 marks our final CPU-only release and ships with a full set of features and improvements, including: ?? The biggest update to Mojo ever – with support for Python 3.12, core language improvements, new standard library APIs, stronger VS Code integration, and virtual development environments based on Conda packaging. ?? Magic, a new standards-based package manager, delivering MAX and Mojo in a unified package that's more than 30% smaller than previous releases. ?? An updated Llama 3.1 pipeline – with a 45% performance boost compared to the 24.4 release - delivering SOTA performance for CPU inference. It's built on the new MAX Driver interface, giving developers more control over the MAX engine and hardware devices. ?? New Python bindings for the MAX graph API, bringing the speed of MAX and Mojo directly to your existing workflows. ? Learn more about this latest release on the Modular blog. It's another step forward on our mission to reinvent how AI technology is developed and deployed into production, and we're excited to share what's coming next! ?? https://lnkd.in/g29Q-p9C

What’s been cooking in the Modular kitchen lately? ???? This month has been packed with ?? technical insights and community contributions: ? We explored linear types with Evan Ovadia, uncovering their potential for solving caching problems, guaranteeing the completion of futures, and more. ? We dove into SIMD parallelism with Ash Vardanian, showing how modern CPUs can unlock GPU-like performance. ?? We learned about NuMojo, game development in Mojo, and updates to community-driven projects like "Mojo By Example." Catch all the recent highlights: https://lnkd.in/gZMwrGTv #MojoLang #Programming

Did you know modern CPUs can act like mini GPUs, performing up to 32 operations in parallel, through single instruction, multiple data (SIMD) parallel processing? Yet most of that power sits idle. Why? In our latest blog, guest writer (and founder of Unum) Ash Vardanian joins Chris Lattner, Daniel Lemire, and Evan Ovadia to explore: ? Why SIMD remains underutilized ? Hard-earned insights from building the SimSIMD library ? Practical solutions to empower developers to harness SIMD performance Learn how this tech is already transforming AI and DBMS software on 100M+ devices: https://lnkd.in/eAk-5fNq #programming #simd

Struggling with inconsistent data or forgotten function calls in your code? In our recent community meeting, Evan Ovadia explored how linear types could be the solution you didn’t know you needed ?? Check out the recording: https://lnkd.in/gpNv5dEv #MojoLang #programming

Do you ?? Python? With MAX 24.5 and our new MAX Graph API, you can run Llama3.1 on your laptop and chat with your docs at state-of-the-art speeds! ?? See our blog to learn how we built a RAG app with MAX, Streamlit, and Python https://lnkd.in/gyzXmGDS

Could linear types be the secret ?? to solving caching problems, guaranteeing the completion of futures, and more? Join us for Tuesday's community meeting, where Evan Ovadia will explore how linear types could benefit the Mojo compiler. RSVP in Discord: https://lnkd.in/gZNmshVT #programming #mojolang

Earlier this fall we introduced Magic ??, our new command line tool with support for package and virtual environment management. Why the change? Learn about our motivations for the revamp, why we think you'll ?? it, and what's coming next in our latest blog post ?? https://lnkd.in/gJiiZSvU

Thanks to @CodeToTheMoons for featuring the MAX ???? Llama 3.1 ?? pipeline in his most recent video on building a text summarizer with Mojo! ?? https://lnkd.in/gyD6BRxW

Missed our latest community meeting? Don’t worry — the full recording is now live on YouTube! https://lnkd.in/gja2rdkw In this meeting, we heard from: ????Helehex on community-led Mojo standard library contributor meetings and the game development community ???Alex Carey on NuMojo, a Mojo library for numerical computing ???Jack Clayton on changes to Mojo’s Writer trait ??Chris Lattner and team on compile-time programming and long-term vision for the standard library #mojolang #programminglanguages #python

Our next community meeting is right around the corner! On Monday at 10 AM PT, we’ll dive into the Mojo game development community, showcase community projects, learn about updates to the Mojo Writer trait, and host a Q&A session. ?? RSVP on Discord: https://lnkd.in/gRyvQ5J3 #programminglanguages #ai #ml