What Is Mojo and How It Relates to Python

Mojo is a programming language developed by Modular Inc., the AI infrastructure company co-founded in 2022 by Chris Lattner and Tim Davis and acquired by Qualcomm in July 2026. It was unveiled on May 2, 2023, with a bold pitch: Python's syntax with the performance of C and direct access to modern hardware such as GPUs. If you are wondering what Mojo is and how it relates to Python, the short answer is that it looks familiar to Python programmers, works alongside Python code, but is a separate, compiled language designed for high-performance AI and systems work.

Origins: Lattner's third language project

Lattner is known for creating LLVM, the Clang compiler and Swift, and for co-creating MLIR, a compiler framework for heterogeneous hardware, while at Google. Modular's goal was to rebuild the AI software stack so the same code could run efficiently on CPUs, GPUs and accelerators from different vendors. The team found that Python was the language AI developers wanted to use, but its performance meant the important code was actually written in C++ and CUDA. Mojo was built to close that gap, and it powers Modular's own MAX inference platform.

What is Mojo and how it relates to Python

At launch, Modular described the long-term goal of Mojo becoming a superset of Python. Over time the company softened that message, describing Mojo instead as a Pythonic language that is a member of the Python family, with full superset compatibility not an immediate priority. The practical relationship looks like this:

• Syntax: indentation-based blocks, def functions and familiar control flow make Mojo easy to read for Python developers.
• Typing and compilation: Mojo is statically typed and compiled ahead of time or just in time through MLIR and LLVM, so typed code runs at native speed.
• Interop: Mojo can import and call Python modules such as NumPy through the CPython runtime, and Modular has added ways to call Mojo code from Python.
• Systems features: structs, traits, compile-time parameters, explicit SIMD types and an ownership model with value semantics, borrowing ideas from Rust and Swift.

A short Mojo example

This snippet defines a typed recursive function and calls it from a main entry point. Mojo's syntax has changed between releases, so check the current documentation before relying on details:

fn fib(n: Int) -> Int:
    if n < 2:
        return n
    return fib(n - 1) + fib(n - 2)

def main():
    for i in range(10):
        print(fib(i))

Where Mojo is used

Mojo's main use today is writing high-performance kernels for AI workloads, especially GPU code, as an alternative to CUDA C++ that can target hardware from more than one vendor. Modular uses it heavily inside its own inference stack, and a community of early adopters experiments with numerical computing, data processing and performance-sensitive Python extensions. Early marketing highlighted dramatic speedups over Python on benchmarks such as the Mandelbrot set; those figures were Modular's own measurements on specific workloads and compare against pure Python, not against NumPy or C.

Milestones

Limitations

Mojo is still young and should be treated as a language that is still evolving. Version 1.0 arrived in August 2026 with a promise of far fewer breaking changes, but the language was changing in breaking ways for years before that, and many Python features, such as full class support with dynamic behavior, are not implemented in the same way. The standard library has been open source since 2024, and in August 2026 Modular also open-sourced the compiler under the Apache 2.0 license with LLVM exceptions, although outside contributions to the compiler and tooling were only just opening up. Platform support has been narrower than Python's, the package ecosystem is small, and the language's direction is closely tied to one company.

Should you learn Mojo?

If you write performance-critical AI code, GPU kernels or numerical routines and you already know Python, Mojo is worth experimenting with. For general application development, data science or web work, stick with Python for now. Knowing what Mojo is and how it relates to Python keeps expectations realistic: it is not a faster drop-in Python interpreter, but a new systems language that speaks Python's dialect and can work alongside it.

Frequently asked questions

Is Mojo a superset of Python?

Not today. Modular originally described becoming a superset as a long-term goal, but later framed Mojo as a Python-family language instead. It can import and use Python libraries, but most Python code will not run unchanged as Mojo code.

Is Mojo open source?

Yes, for the language itself. The Mojo standard library was open-sourced in 2024, and the compiler and tooling followed in August 2026 under the Apache 2.0 license with LLVM exceptions. Modular's MAX inference framework is not part of that release.

Who created Mojo?

Mojo was created at Modular, the company co-founded by Chris Lattner and Tim Davis. Lattner previously created LLVM, Clang and Swift, and co-created the MLIR compiler framework.