A language becomes an ecosystem when its community starts building the libraries everyone else relies on. That’s happening with Mojo and MAX, and we want to support more of it. Meet the first Modular Community Grant cohort: six community-built Mojo and MAX libraries, funded by Modular, covering reinforcement learning, GPU physics, and developer tools.

Nurturing the Modular ecosystem: the Modular Community Grant Program

Source: https://www.modular.com/blog/nurturing-the-modular-ecosystem-the-modular-community-grant-program?utm_source=x We openeda call for community project proposalsearlier this year, offering small grants to support Mojo and MAX libraries forming the foundation of an ecosystem. We heard from a wide range of contributors and received many more strong proposals than we are able to support.

Here are the projects we’re supporting right now:

noeira

Denis Laboureyras (@denislabs)

Created in early 2026,noeirais an open-source reinforcement learning (RL) framework written entirely in Mojo. The project currently supports 40+ RL algorithms, GPU-accelerated deep RL, custom 2D/3D physics engines, native arcade game engines and SDL3 rendering.

Born of frustration with performance cliffs in the Python reinforcement learning stack, the long-term goal of this project is to provide a fully end-to-end RL framework where environments, agents and training all run on the GPU in pure Mojo.

In the near term, the project aims to publish rigorous benchmarks against established RL stacks (Gymnasium,Stable Baselines3,PyTorch) as a way of quantifying performance gains obtained by taking a pure Mojo approach.

Watch Denis’ community meeting presentation here.

bajo

Jean-Gabriel Simard (@jgsimard)

A parallel physics simulator,bajois designed to run thousands of RL/robotics training environments simultaneously on a single GPU. The project has already shipped a state of the art GPU sorting implementation (Morton encoding + BVH construction) as its compute foundation.

The six-month plan is to package that sorting layer, implement performant parallel collision detection and rigid-body dynamics. This will then be scaled to a working multi-thousand-environment prototype and benchmarked against NVIDIA’sWarpandMadrona.

Watch Jean-Gabriel’s community meeting presentation here.

Fast Fourier Transform

Martin Vuyk Loperena (@martinvuyk)

A Fast Fourier Transform (FFT) implementation that’s already approaching state of the art performance on both CPU and GPU, competitive withFFTWandcuFFTrespectively. Started at a hackathon in mid-2025, the project has been continuously evolving since then.

Near term goals include adding half-sized transform support (hC2R, R2hC) and pushing for outright performance wins on the R2R, C2R and R2C transforms already supported. If the remaining gap can be closed, MAX’scuFFTdependency could be completely removed.

Catch the November community meeting for Martin’s presentation on his FFT implementation.

EmberJson

Brian Grenier (@bgreni)

EmberJsonis the most widely used JSON parsing library in Mojo today. The package leans heavily on Mojo’s reflection capabilities and supports both structured and unstructured parsing.

The roadmap points toward a general serialization framework in the spirit of Rust’sSerde, plus continued work on Python interoperability. Early benchmarks already show it out-pacing Python’s built-in JSON handling despite interop overhead.

Join our October community meeting to hear more from Brian about the progress he’s made on this project.

floki

Mikhail Tavarez (@thatstoasty)

flokiis an easy to use Mojo HTTP client built onlibcurlbindings. The project already supports many of the features users of Python’srequestsmodule know and love. Development is now focused on providing ergonomic access to advancedlibcurlconfiguration and simplifying the implementation of the underlying bindings.

Watch Mikhail’s community meeting presentation here.

momanim

Josiah Laivin (@josiahls)

momanimis a 2D animation framework for Mojo inspired byManim. The project is aimed at using CPU and GPU acceleration to dramatically reduce rendering times and targets an API close enough to Manim’s that LLM-assisted Python-to-Mojo conversion is realistic. The project is in its infancy: basic image/video APIs and shape rendering are already in place and a 3D rendering API is next on the roadmap.

Tune into the October community meeting to hear from Josiah about momanim.

The shape of things to come

Taken together, this cohort is a snapshot of where the Modular community is headed as Mojo and MAX continue to develop: foundational infrastructure and exciting applications. We’re thankful to everyone in the community for pushing the platform forward and we can’t wait to see what you’ll do next.