Modular · 2026-05-07 · major
Mojo 1.0 Beta — Modular's Python-Performance AI Language Hits Feature Freeze, Compiler Open-Source Coming at Final 1.0
Chris Lattner's team shipped the first feature-complete Mojo 1.0 beta, plus MAX 26.3 with a distributed-aware Tensor, AMD MI250X / NVIDIA B300 / Apple M5 GPU backends, and Wan 2.2 video generation.
Chris Lattner's Python-on-the-front, GPU-on-the-back AI language reaches feature-complete with new AMD, NVIDIA, and Apple Silicon backends.
Key specs
| Version | 1.0.0b1 |
|---|---|
| Platform release | Modular 26.3 |
| Stability promise | 1.x APIs stable across releases |
What is it?
Mojo is Modular's high-performance systems language that targets a Python-like syntax with C++/Rust-class speed for CPU and GPU code. Beta 1 is what the team considers a feature-complete Mojo 1.0; the months until final 1.0 are about polish, semantic versioning, and stable APIs that won't break across the 1.x series. MAX 26.3, Modular's inference platform, ships alongside.
How does it work?
Mojo compiles MLIR through a custom toolchain to CPU and GPU without a runtime tax. The 1.0 beta unifies def/fn, formalizes safe closures with a new capturing syntax, makes UnsafePointer non-null by default, swaps LayoutTensor for a TileTensor type with compile-time GPU layouts, and lights up Apple Metal print()/dynamic memory plus M5 matrix ops, AMD MI250X, and NVIDIA B300. MAX 26.3 adds a distributed-aware Tensor with multi-GPU compilation, the Wan 2.2 video model, FP8/NVFP4/MXFP4 quantization via QuantConfig, and tuned NVFP4 grouped matmul.
Why does it matter?
Mojo has been one of the loudest bets on a CUDA-class AI language with a Python on-ramp. A real 1.0 milestone — with Modular publicly committing to open-source the compiler when 1.0 lands later this year — pulls Mojo from 'interesting research preview' to a target real teams can plan production AI infrastructure around. The expanded GPU backends also make MAX a more credible cross-vendor inference engine.
Who is it for?
AI infrastructure engineers, GPU kernel authors, performance-sensitive ML teams.
Try it
uv pip install --upgrade modular