Signals
Bun 1.4 Rewrites the Runtime in Rust: 5× Lower Idle CPU and Up to 35% Less Memory
Bun 1.4 rewrites its runtime core from Zig to Rust. Bun reports 5× lower idle CPU, up to 35% lower memory use, and 50% faster Linux startup, with workload-specific gains including lower Claude Code CPU. Existing Bun users should upgrade on a branch and measure their real CI, agent, or server workload; stable Node teams do not need to migrate.
What Changed
- Bun 1.4 rewrites the runtime core from Zig to Rust. Bun says every release benchmark matches or beats 1.3, while the Linux x64 binary is about 20% smaller.
- Bun reports 5× lower idle CPU on a small hello-world process and roughly 2× lower production CPU for Claude Code, where p99 fell from 24% to 10% and p50 from 5.8% to 2.5%.
- Bun summarizes the release as using up to 35% less memory. Its detailed HTTP-server benchmarks vary by workload: Next.js peak memory fell from 397 MB to 285 MB, while Express fell from 169 MB to 92 MB.
- Startup also moved materially: Bun reports 50% faster startup on Linux, with its hello-world benchmark dropping from 10.9 ms in Bun 1.3 to 5.1 ms in Bun 1.4.
Who Should Care
- Teams already running Bun in CI, local development, or agent-heavy repositories. Repeated startup and idle overhead compound across worktrees, test runs, and retries, so benchmark 1.4 against the pinned version.
- Bun-backed server workloads where memory density or idle CPU matters. Compare RSS and CPU before and after on the same traffic pattern instead of assuming the vendor benchmark maps directly to your service.
- Teams deciding whether Bun is mature enough to standardize on. The Rust rewrite, smaller binary, lower runtime overhead, and stronger Node compatibility are useful signals even if you do not change stacks yet.
Who Should Not Care
- Stable Node.js applications with no current startup, memory, CI, or runtime-cost problem. A good Bun benchmark is not by itself a reason to absorb migration risk.
- Teams looking for a universal 5× speedup. The 5× figure is specifically Bun's idle-CPU result for a small process, while memory and CPU gains vary materially by workload.
- Anyone who cannot benchmark the upgrade against the same application and workload. Treat the published numbers as vendor evidence to test, not guaranteed production results.
The Verdict
If I already had Bun in the repo, I would upgrade 1.4 on a branch and replay the exact workload I care about: the same test suite, build, agent loop, and representative server process. Measure wall-clock, peak RSS, and CPU, then compare it with the current pinned version. The interesting part is not one isolated benchmark. It is whether lower startup, idle CPU, and memory compound across the dozens or hundreds of times your agents and CI start the runtime every day. I would not migrate a stable Node app on these numbers alone, but Bun 1.4 makes the case for re-running that comparison much stronger.
Take This To Your Agent
Copy a ready-to-paste investigation handoff that asks your coding agent to check this update against your product and recommend what to do next.