Vitest vs Jest vs node:test in 2026: Benchmarked on a Real TypeScript Project

In November 2026, a TypeScript test suite was benchmarked on a Linux VM with 8 vCPUs and 16 GB RAM using three test runners: Vitest 5.1, Jest 30.5.2 with SWC, and Node's built-in `node:test`. The tests consisted of 39 files and 183 tests that covered a NestJS-style orders service, HTTP API, and MCP server.

The benchmark revealed that with default settings, Jest with SWC was the fastest full run, while Vitest was the slowest. However, when isolation settings were adjusted, the ranking changed. Without isolation, Vitest and Node's `node:test` were both faster than Jest.

The study also examined different configurations, such as using threads or vmThreads for Vitest, and type-checking with ts-jest. It found that isolation was the most significant factor affecting performance, with Vitest's default forks pool running each test file in its own child process and Jest's separate VM contexts for each file. Disabling isolation led to significant improvements in both Vitest and Node's `node:test` times.

Additionally, the benchmark showed that as the test suite grew in size (to 156 files and 732 tests), the performance gap between the runners widened. Vitest's performance remained relatively stable, while Jest's time increased more noticeably. The study concluded that while the exact timings may vary depending on the machine and environment, the overall trend would likely remain consistent.