Bun 1.4.4: 2,025 Files Formatted in 2.6 Seconds, With bun check, bun install, and Bun.SQL in the Same Patch
A runtime can replace Prettier, lint-staged, tsc, and parts of pnpm/Yarn in one upgrade, which simplifies toolchains but leaves the 0.9% compatibility gap as a user-side risk. Windows and other non-Linux users should reproduce the official benchmarks before assuming the same speedups.
Bun 1.4.4 puts four toolchain upgrades behind one version bump. On Prettier's own repository, bun fmt formatted 2,616 files in 59.7 ms with 48 MB of peak memory, compared with 17.1 seconds and 459 MB for Prettier 3.9.10. Across vitest, date-fns, monkeytype, and h3, bun fmt measured 234 to 425 times faster than Prettier 3.9.9, and it beat oxfmt 0.74.0 by 31 times on fumadocs, 6.2 times on vite, and 3.0 times on RSSHub. On a Windows machine, 2,025 date-fns files took 2,643 ms on the first pass and 916 ms on the second, with user CPU time of only 0.077 seconds.
Migration tooling is built in. bun fmt --doctor reads existing config like .prettierrc and lists which packages can be removed; --check gives CI an exit code 1 when files need formatting; --staged replaces lint-staged in pre-commit hooks. Compatibility covers 9,648 of 9,731 Prettier test cases and 94.6% alignment with oxfmt across 330 open-source repos, with four intentional differences. bun check cuts a 4,000-project monorepo batch check from 13.65 seconds to 0.87 seconds and drops peak memory from 1,987 MB to 872 MB. bun install adds exponential backoff retries, Yarn Berry lockfile migration, musl-aware package selection, and a minimumReleaseAge option for blocking freshly published versions.
Bun.SQL reduces PostgreSQL client instruction counts by 21 to 48 percent in the benchmarked query shapes and adds query.cancel() plus COPY streaming. The upgrade advice is tiered: bun fmt can be switched immediately, bun check is worth testing against tsc, and Bun.SQL should wait until a real slow-query case exists.
The 99.1% Prettier test pass rate leaves 83 failing cases, and any one of those can break a real file; --doctor and --check exist because compatibility stops short of complete.
Because bun fmt is a port of oxc_formatter, which is itself a port of Biome's formatter, two layers of translation sit between its output and Prettier; the 99.1% pass rate is a chain of assumptions about how faithfully each port preserved the previous behavior.
The patch release consolidates formatter, hook, type-check, package-manager, and SQL-client duties into a single runtime, which means one upgrade changes an entire toolchain and one edge case can affect all of it.
Official benchmarks on Linux and macOS are not directly comparable to the 2.6-second Windows result, so cross-platform filesystem and thread behavior matters more than formatter CPU time in real usage.