NewsDeveloper Tools

Astral’s ty Type Checker Beta: 80x Faster Than Pyright

Astral released the beta version of ty on December 16, 2025—a Rust-powered Python type checker that’s 10-60x faster than mypy and Pyright. This is Astral’s third major Python tool built in Rust, following Ruff (linter) and uv (package manager). The beta announcement signals production-readiness, with Astral already using ty exclusively in their own projects.

The Type Checking Adoption Gap

Here’s the problem: 73% of Python developers use type hints in production code, but only 41% actually run type checkers in CI. Why the gap? Speed. When mypy takes 36 seconds just to import PyTorch, or 15+ minutes to check 4 million lines at Dropbox, developers disable it to keep builds moving.

A 2024 survey by Meta, JetBrains, and Microsoft identified “slow performance of type checkers like Mypy” as the second-biggest pain point for over 1,000 Python developers. ty was built specifically to close this gap.

Performance That Changes Behavior

ty isn’t just faster—it’s fast enough to fundamentally change how developers work. In the PyTorch codebase, after editing a core file, ty recomputes diagnostics in 4.7 milliseconds. Meanwhile, Pyright takes 370ms. That’s 80x faster. For the home-assistant project, ty completes a full type check in 2.19 seconds versus Pyright’s 19.62 seconds.

The secret is ty’s incremental architecture. Instead of rechecking everything when you edit a function, ty selectively recomputes only what’s necessary. Consequently, in an editor, this means type errors appear almost instantly—no lag, no coffee breaks.

Astral’s Proven Formula

This is Astral’s third time executing the same playbook: identify a slow Python tool, rewrite it in Rust, achieve massive adoption. Ruff replaced Flake8 and Black with 10-100x speedups and was adopted by FastAPI and Airflow within months. Similarly, uv consolidated pip, pip-tools, pipx, poetry, and pyenv into one tool.

The pattern works because Astral doesn’t just make things faster—they maintain compatibility while adding features. ty isn’t “mypy but in Rust.” Instead, it’s a complete language server with auto-import suggestions, go-to-definition, symbol rename, code completions, and error messages inspired by Rust’s legendary diagnostics.

Beta Means Production-Ready

Astral is already dogfooding ty across their codebase and recommends it for “motivated production users.” The beta tag in 2025 doesn’t mean experimental—it means they’re confident enough to use it internally while refining edge cases based on community feedback.

Installation takes one command: uvx ty check. Furthermore, there’s also a VS Code extension, plus support for Neovim, Zed, and PyCharm 2025.3+. Four releases shipped between December 16-20 (v0.0.2 through v0.0.5), showing active development and quick iteration.

The Rust Revolution in Python Tooling

ty isn’t happening in isolation. The entire Python tooling ecosystem is shifting to Rust. Competitors like Pyrefly are also Rust-based, though still 500x slower than ty. The reason is simple: Python’s flexibility makes it hard to write fast Python tools in Python.

Astral’s long-term vision ties ty into a broader ecosystem. Think uv for package management, Ruff for linting and formatting, ty for type checking—all sharing semantic understanding. Future capabilities include dead code elimination, unused dependency detection, and type-aware linting that works across the entire toolchain.

What Happens Next

The roadmap prioritizes stability, completing Python’s typing specification, and adding first-class support for Pydantic and Django. A stable 1.0 release is targeted for 2026. In the meantime, developers can try ty today via the GitHub repo or online playground at play.ty.dev.

Python’s type checking problem—slow tools that discourage adoption—is now solved. The question is how fast the ecosystem migrates.

ByteBot
I am a playful and cute mascot inspired by computer programming. I have a rectangular body with a smiling face and buttons for eyes. My mission is to simplify complex tech concepts, breaking them down into byte-sized and easily digestible information.

    You may also like

    Leave a reply

    Your email address will not be published. Required fields are marked *

    More in:News