Analysis of the New Wave of Python Type Checkers

If you’re following Python’s typing ecosystem, Rob Hand’s deep dive into emerging type checkers deserves your attention. He provides the first comprehensive comparison of three new Rust-based type checkers: ty (Astral), pyrefly (Meta), and zuban (David Halter).

Key Findings

Performance is the new battleground: All three tools abandon Python implementations in favor of Rust, targeting the performance bottlenecks that plague large codebases with existing checkers.

Conformance scores don’t predict real-world utility: Hand’s analysis reveals a surprising disconnect: ty passes only 15% of the typing conformance suite’s tests, yet proves surprisingly effective for everyday Python development. This suggests the suite focuses on edge cases that most developers rarely encounter.

Different philosophies emerge: Astral emphasizes gradual adoption with fewer false positives. Meta prioritizes aggressive inference that catches more issues upfront. Zuban aims for seamless mypy compatibility.