Mojo 1.0 Open Source Release and Modular Platform Updates

Mojo 1.0 is Now Fully Open Source

Mojo 1.0 has been released as a fully open-source language under the Apache 2.0 license. This release includes the compiler and all associated tooling, allowing developers to extend the language, port it to new platforms, and build applications without proprietary restrictions. This follows a progressive opening of the Modular stack, which previously included the Mojo standard library in 2024 and MAX kernels in 2025.

Stability and Design

Mojo 1.0 introduces a stability guarantee, ensuring that code written for this version will not break in future updates. The language is designed to provide a production-ready foundation for AI development, combining a Python-like syntax with systems-level performance.

Community discussion highlights that Mojo is built on top of MLIR (Multi-Level Intermediate Representation), which allows compiler optimizations to be implemented as libraries rather than within the compiler itself. Some developers describe Mojo as a "fancy MLIR wrapper" that implements semantics inspired by C++ but presented with a Python-like skin.

Native Windows Support

Modular is collaborating with the Microsoft Windows team to bring native Windows support to Mojo. While Mojo previously supported macOS and Linux (and Windows via WSL), native integration aims to make the language accessible to millions of Windows developers working across systems and AI workloads.

Modular Cloud General Availability

Modular Cloud is now generally available at console.modular.com. It serves as a production service that abstracts the complexity of deploying and optimizing models across heterogeneous infrastructure.

Service Offerings

  • Shared Endpoints: OpenAI-compatible endpoints with pay-per-token pricing.
  • Dedicated Deployments: Isolated instances running on either Modular's compute or customer-owned infrastructure.

Production Case Study: MiniMax

MiniMax utilizes Modular Cloud to run its M3 model, serving billions of tokens per minute. The deployment required deep optimization across the stack, including the implementation of M3 natively in MAX and the development of specialized kernels for MiniMax Sparse Attention (MSA), a novel architecture that reduces compute requirements as context scales for its 1M-token context window.

Expanded Hardware Support for Modular Platform

The Modular Platform has expanded its support beyond NVIDIA and AMD GPUs to include custom AI accelerators. This allows developers to author a model once and deploy it across different hardware architectures without rebuilding the software stack.

Supported Accelerators

  • AWS Trainium
  • Google TPUs
  • Qualcomm Cloud AI 100 Ultra
  • Qualcomm Dragonfly

Modular reports a 10x reduction in the engineering effort required to bring up these new platforms compared to traditional methods. For example, HTEC engineers independently brought up Google TPU support on the Modular Platform within a few months.

The MAX Framework and Ecosystem Alliance

MAX is the core foundation of the Modular Platform, designed to allow hardware vendors to compete on silicon merits rather than software stack lock-in.

Licensing and Open Alliance

  • Source Availability: The MAX license no longer contains device usage restrictions and is now source-available.
  • Alliance Program: Modular is establishing an industry alliance program involving hardware vendors, model providers, cloud companies, and data-center operators to integrate and optimize MAX for various platforms.

Community Perspectives and Critiques

While the open-sourcing of Mojo 1.0 has been welcomed, some community members have expressed skepticism regarding the timing and adoption of the language.

"The long time it took them to open source I think burnt a lot of the initial traction."

Other critics have questioned the strategic intent behind the move following Qualcomm's acquisition of Modular, with some suggesting it is a strategic move to challenge NVIDIA's dominance in the GPU market. Conversely, proponents argue that Mojo's approach to numerics and static typing makes it a compelling alternative to Rust for AI-centric systems programming.

Sources

Related

  • Dispatch
    Mojo 1.0 Release Notes

    Modular has released Mojo 1.0, establishing a stable, production-ready foundation for its systems programming language designed for AI infrastructure.

  • Dispatch
    Mojo Language Open Source Release

    Modular has open-sourced the Mojo compiler and toolchain under the Apache 2.0 license, enabling developers to build and customize the systems language for AI accelerators.

  • Dispatch
    Mojo 1.0 Beta: Bridging the Gap Between Python Productivity and C++ Performance

    Mojo enters beta as a compiled, statically-typed language designed for AI and systems programming, promising the syntax of Python with the performance of C++.

  • Project
    modular/modular

    A unified platform for AI development and deployment featuring the Mojo programming language and the MAX framework for optimized model inference.

  • Dispatch
    Qualcomm Acquires Modular

    Qualcomm has acquired AI startup Modular in a deal reportedly valued at $4 billion to integrate Modular's compiler stack and Mojo language into its hardware ecosystem.