Mojo
A programming language built for AI work that looks like Python but compiles to fast code for CPUs and GPUs.
AI work has an awkward split down the middle. Researchers write in Python because it is comfortable, and the parts that actually need to be fast get rewritten in C++ or CUDA, usually by different people. Mojo, from the company Modular, tries to close that gap: familiar Python-like syntax on the surface, but compiled down to code that runs well on processors, graphics cards and AI accelerators.
The project is led by Chris Lattner, whose earlier work includes the LLVM compiler toolkit and Apple’s Swift language, which is why it was taken seriously from the start. Mojo reached version 1.0 in August 2026, meaning the language is now considered stable enough to build long-lived projects on. Worth knowing: the standard library is open source, but the compiler itself is not yet, though the company has promised to open it during 2026.