March 16, 2026
Type wars: Pyright vs the world
Comparing Python Type Checkers: Typing Spec Conformance
Python’s type checker war: one tool tops the charts, but devs are fighting about whether any of this even makes sense
TLDR: A new scoreboard shows which Python tools best follow the official typing rules, with Pyright leading and mypy lagging. Commenters, however, are split between mocking Python’s “decorative” types, praising user‑friendly newcomers like ty, and arguing whether strict rule‑following even matters for everyday code.
Python just got its own scoreboard for “who follows the rules best,” and the type checker Olympics have a clear winner: Pyright at the top, with the long‑time favorite mypy choking at barely over half the tests. But the real action isn’t in the numbers – it’s in the comments section, where devs are openly roasting Python’s whole type system.
One commenter basically shrugs at the whole article and says it’s just a pretty write‑up of the official results, while another goes nuclear, calling Python’s type hints “utter idiocy” because you can label something as a number and then happily assign it a string without the language complaining. That line had half the thread nodding in pained agreement. Meanwhile, fans of a newer tool called ty show up like a street team, insisting it’s fast, friendly, and doesn’t annoy coworkers who never wanted types in the first place.
Others are asking the nerdier questions, like whether anyone can please invent a tool that checks the shapes and sizes of AI data arrays before everything explodes at runtime. And in the background, a quieter debate simmers: should we chase perfect rule‑following, or just tools that catch common bugs without driving people crazy? The scoreboards say one thing; the crowd definitely says another.
Key Points
- •Python’s typing semantics have been consolidated into a formal specification with a conformance test suite.
- •The test suite encodes expected errors and non-errors to detect false positives and false negatives.
- •Conformance results (early March 2026) show Pyright at 136/139 (97.8%), Zuban 134/139 (96.4%), Pyrefly 122/139 (87.8%), mypy 81/139 (58.3%), and ty 74/139 (53.2%).
- •A public dashboard aggregates conformance outcomes; figures are evolving, especially for beta tools like Zuban, Pyrefly, and ty.
- •Lower conformance can force developers to restructure code or apply workarounds; Pyrefly claims support for major features and aims to close remaining gaps.