Mojo Language

This release graduates the eager-style Tensor and Module APIs to stable status, simplifying GenAI model building, and significantly expands Apple silicon GPU support for basic MAX graphs. It also introduces several powerful new features to the Mojo language, including compile-time reflection and linear types.

Release 25.7 introduces a fully open-source MAX Python API and an experimental PyTorch-like Model API, alongside significant performance improvements and expanded Mojo features including better Apple silicon GPU support.

Release 25.6 introduces significant compute layer unification in MAX, adding support for NVIDIA Blackwell and AMD MI355X throughput, along with initial Apple Silicon GPU support for Mojo and a standalone `pip install mojo` package.

Modular Platform 25.5 introduces Large Scale Batch Inference, the open source MAX Graph API, and expanded PyTorch operator support, alongside optimized deployment tools.

Modular Platform 25.4 introduces full support for AMD GPUs, significantly enhancing hardware flexibility and democratizing access to high-performance AI compute.

Modular Platform 25.3 unifies installation via "pip install modular", open-sources MAX components, and adds Google Colab support, lowering barriers for community engagement.

MAX 25.2 introduces significant enhancements for large-scale AI deployment, including comprehensive NVIDIA Hopper support, multi-GPU tensor parallelism, and new memory optimization features like GPTQ quantization.

Release 25.1 significantly improves developer experience for Agentic and LLM workflows, introduces a new nightly release model with a GPU programming interface, and launches MAX Builds.