The fastpybase64 package hit PyPI with a benchmark that decodes base64 at 10 GB/s on an Intel Xeon Gold 6548N core using Python 3.14.6, four times faster than pybase64 and 27× faster than the built‑in base64 module.
Measured throughput
The author measured single‑threaded decode speed on random data broken into 76‑character MIME lines, the format used for email attachments. Results on a 1 MiB sample were:
- stdlib base64: 0.49 GB/s
- pybase64 1.5: 3.2 GB/s
- fastpybase64 0.1.0: 13.1 GB/s
These numbers translate to roughly 10 GB/s of decoded payload on the hardware used, confirming the claim of “almost the speed of a memory copy.”
Drop‑in replacement
fastpybase64 mirrors the API of the standard library. After pip install fastpybase64 you can import fastpybase64 as base64 and call b64encode/b64decode exactly as before. The module supports Linux, macOS, and Windows and runs on Python 3.10‑3.14, including free‑threaded Python 3.14.
What the announcement doesn’t cover
The post does not list any pricing model – the package is free on PyPI. It also does not discuss multi‑core scaling; the benchmark is pinned to a single core, so you won’t see linear gains on many cores without explicit parallelism. Finally, the performance numbers rely on a high‑end Xeon CPU; lower‑end hardware may see smaller speedups.
When to try it
If your service ingests large volumes of base64‑encoded data – for example, processing email attachments, data URLs, or binary blobs in JSON – swapping the stdlib module for fastpybase64 can reduce CPU time dramatically. Test it on your target hardware and Python version before rolling out, especially if you run on older CPUs or need multi‑threaded throughput.
Read the full blog post for benchmark details and source code.
