Modular's Mojo programming language hits 1.0 milestone

Modular s Mojo attained its 1.0 milestone on Tuesday, marking a stable and production-ready general-purpose programming language. The language's main objective is to unify the AI software stack, enabling developers to target diverse GPUs, CPUs, ASICs, and other systems that run AI workloads. In collaboration with MAX inference framework, Mojo aims to simplify AI software, systems, and tool development, without relying on vendor-specific platforms such as Nvidia's CUDA or AMD's ROCm.

CUDA, Nvidia's proprietary GPU programming platform, is often preferred by AI-oriented tech companies seeking an open-source alternative for optimization on their specific hardware. Chris Lattner, known for creating LLVM, Clang, Swift, and MLIR, has been developing Mojo since 2023 through his startup Modular, which was acquired by Qualcomm in June.

Mojo strives to combine the accessibility of Python with Rust-like memory safety and system programming capabilities, often found in languages like C, C++, and CUDA. The 1.0 release introduces new features like Python-style lambda syntax, an improved Language Server Protocol for code editors, and Mojo AI skills. The language's standard library is under the Apache License v2.0 with LLVM Exceptions.

Modular plans to open-source the compiler this year, with the language website hinting at a potential release at Modcon next week. During a streamed Q&A session, Lattner addressed concerns about Qualcomm's acquisition of Modular, stating that Nvidia and AMD remain strong open-source contributors, and he doesn't anticipate any issues with the platform's vendor-neutrality.