The Modular universe is buzzing! From next-level community projects to recognition across the AI and developer space, here’s the latest from our growing ecosystem.
Community Innovations
The Modular community never fails to impress us with its creativity and innovation. Developers from around the world share projects that push the boundaries of what’s possible with MAX and Mojo, from tools that make AI development easier to experiments that explore new frontiers in performance. Discover what the Modular community has been building:
- In our recent community meeting, we heard from Martin Vuyk on his generic FFT implementation in Mojo, which delivered up to 18× faster performance than cuFFT in his benchmarks, and from Gabriel de Marmiesse on his MAX backend for PyTorch. Catch the full details in our October Community Meeting recording.
- mewmix’s project is a pure Mojo implementation of Keccak-256, complete with a self-contained test harness and reproducible test vectors.
- Photon created SciJo, a high-performance scientific computing library for Mojo, providing SciPy-like functionality with the speed and efficiency of native Mojo code.
- Laranzu created a Julia fractal image generator in Mojo for both CPU and GPU, providing an accessible example for Python programmers and demonstrating Mojo’s impressive performance compared to C++ and Python.
- Sören developed duckdb.mojo, providing Mojo bindings that let developers tap into DuckDB’s high‑performance analytical database.
- Constantin G created Butterfly, a new Mojo-based qubit simulator that already delivers 10–15x faster performance than Qiskit while showcasing how elegantly Mojo handles core quantum operations.
- Ash Vardanian’s Scaling Elections explores how the computationally expensive Schulze voting method can be massively accelerated using GPUs and Mojo — enabling fairer ranked‑choice elections to scale efficiently.
- spluta created MMMAudio, a creative-coding audio environment that uses Python for control and Mojo for audio, enabling both high-level instrument building and low-level DSP in one unified workflow.
💡 We love seeing what you create! Share your MAX or Mojo projects in the forum under the Community Showcase category so we can highlight them in the next edition.
Modular Making Waves
Modular continues to make waves in AI and high-performance computing, from leadership appearances to benchmarks demonstrating MAX and Mojo’s capabilities.
- Independent benchmarking by Meta AI recently showcased at Triton Conf shows Mojo matching or outperforming CUDA and significantly outpacing Triton and TileLang DSLs, highlighting Modular’s high-performance capabilities.
- Inworld TTS 1 MAX, powered by Modular Platform, delivers top-tier performance on the Artificial Analysis Speech Arena Leaderboard, surpassing benchmarks set by MiniMax’s Speech-02 series and OpenAI’s TTS-1 series. Modular is proud to power the infrastructure behind this achievement.
- Chris Lattner, Co-Founder & CEO of Modular, shares his Swift-to-Mojo journey 🔥, his vision for bridging AI compute across CPUs and GPUs, and how AI can elevate, not replace, software craftsmanship. Featured on The Pragmatic Engineer and in a conversation with fast.ai co-founder Jeremy Howard
- Tim Davis, President & Co-Founder of Modular, presented in a webinar on building scalable AI systems, highlighting how Modular, AMD, Supermicro, and TensorWave accelerate AI workloads on liquid-cooled systems powered by AMD Instinct MI350 GPUs. Watch replay
- Abdul Dakkak, Senior Director at Modular, showcased Modular Cloud, MAX, and Mojo in a talk on Speed of Light Inference with NVIDIA and AMD GPUs, demonstrating how Modular scales AI workloads efficiently while reducing TCO. Watch here
Open-Source Contributions
If you’ve recently had your first PR merged, message the Modular team in the forum to claim your epic Modular swag! Check out the recently merged contributions from our amazing community members:
- bgreni [1][2]
- c-pozzi [1]
- christoph-schlumpf [1][2]
- cudawarped [1]
- fadend [1]
- hardikkgupta [1]
- johnny19436 [1]
- josetorrs [1]
- josiahls [1][2]
- martinvuyk [1][2][3][4][5][6][7][8][9][10][11][12][13][14][15][16][17][18]
- MaxMeyberg [1]
- mcgrof [1]
- Monal-Patel [1]
- msaelices [1][2][3][4][5][6][7][8]
- Rutzmoser [1][2]
- skrript [1][2][3][4]
- soraros [1][2][3][4][5][6][7][8][9][10][11][12][13][14]
- ThomasMader [1]
- winding-lines [1]
- YichengDWu [1][2]
Modular News & Events: Stay Connected
Here’s what’s new from the Modular team and upcoming events for the community:
- If you’re local to the Bay Area, join us at Modular HQ in Los Altos, CA for our December Meetup! We’ll be doing a deep dive into the MAX platform. Save your spot.
- New YouTube Series! Learning GPU programming through Mojo puzzles - on an Apple M4! No expensive data center GPUs needed. No CUDA C++ complexity. Just Python-like syntax with systems performance. Check out the first video!
- 🎥 Catch up on all past community meetings anytime — the “Modular Community Meetings” playlist has talks, demos, community projects, and discussions with the Modular team.
- Curious how Mojo achieves its GPU performance? This Modular Tech Talk covers the architecture, compiler flow, and open-source kernels behind it, and how Mojo targets CPUs and GPUs in one model.
- 🚀Modular Platform 25.7 is here! With a fully open MAX Python API, next-gen modeling API, expanded NVIDIA Grace support, and safer, faster Mojo GPU programming, developers can spend less time on infrastructure and more time advancing AI. Read more in our announcement blog post!
- From the PyTorch Conference to the LLVM Developers’ Meeting, we noticed a consistent theme: engineers across every layer of the AI infrastructure stack are tackling the same problems — just from different angles. Check out our blog post where we share what we learned from both events and what it reveals about the state of AI innovation in 2025.
