Vercel precompiles Python at build time, cuts cold starts 54%

Vercel now compiles Python to bytecode at build time, not first import, cutting median cold starts from 2.8 seconds to 1.3, per the changelog. Cold start has been the practical reason teams pick Node over Python for edge functions. A 54% cut in that penalty changes the calculation for anyone running FastAPI or Flask.

The old path parsed and compiled Python source at first request after an idle period, adding seconds before a line of application code ran. Node's cold path skips most of that cost, which is why teams defaulted to it for latency-sensitive endpoints even when Python fit the workload better.

Python still starts an interpreter and still costs more than Node's cold path. The build step narrows that gap without closing it. This is a packaging change on Vercel's side, not a shift in how CPython itself starts up. No code changes are required to see the improvement.

If you're running FastAPI or Flask on Vercel, this is a redeploy, not a migration.