1. Install uv
   • Unix: curl -LsSf https://astral.sh/uv/install.sh | sh
   • Windows: powershell -c "irm https://astral.sh/uv/install.ps1 | iex"
2. Build ty: cargo build --bin ty --release
3. cd into the benchmark directory: cd scripts/ty_benchmark
4. Install npm 11.10.0 or newer, which supports the dependency cooldown in .npmrc
5. Install Pyright: npm ci --ignore-scripts
6. Run benchmarks: uv run benchmark

Requires hyperfine 1.20 or newer.

Run with:

uv run --python 3.14 benchmark

Measures how long it takes to type check a project without a pre-existing cache.

You can run the benchmark with --single-threaded to measure the check time when using a single thread only.

Run with:

uv run --python 3.14 benchmark --warm

Measures how long it takes to recheck a project if there were no changes.

Note: Of the benchmarked type checkers, only mypy supports caching.

Measures how long it takes for a newly started LSP to return the diagnostics for the files open in the editor.

Run with:

uv run --python 3.14 pytest src/benchmark/test_lsp_diagnostics.py::test_fetch_diagnostics

Note: Use -v -s to see the set of diagnostics returned by each type checker.

Measure how long it takes to recheck all open files after making a single change in a file.

Run with:

uv run --python 3.14 pytest src/benchmark/test_lsp_diagnostics.py::test_incremental_edit

Note: This benchmark uses pull diagnostics for type checkers that support this operation (ty), and falls back to publish diagnostics otherwise (Pyright, Pyrefly).

The tested type checkers implement Python's type system to varying degrees and some projects only successfully pass type checking using a specific type checker. We benchmark against the latest version of each type checker, but some projects may use another version in practice, leading to different diagnostic results between benchmarking and reality.

The benchmark script supports snapshotting the results when running with --snapshot and --accept. The goal of those snapshots is to catch accidental regressions. For example, if a project adds new dependencies that we fail to install. They are not intended as a testing tool. E.g. the snapshot runner doesn't account for platform differences so that you might see differences when running the snapshots on your machine.