Hey everyone! After sharing some Node.js benchmarks, I received a lot of requests for a similar analysis on Python, so I decided to run benchmarks from Python versions 3.9 through 3.14. Here's a quick look at the results:
|Benchmark|3.9.25|3.10.19|3.11.14|3.12.12|3.13.11|3.14.2|
|:-|:-|:-|:-|:-|:-|:-|
|HTTP GET throughput (MB/s)|9.2|9.5|11.0|10.6|10.6|10.6|
|json.loads (ops/s)|63,349|64,791|59,948|56,649|57,861|53,587|
|json.dumps (ops/s)|29,301|30,185|30,443|32,158|31,780|31,957|
|SHA-256 throughput (MB/s)|3,203.5|3,197.6|3,207.1|3,201.7|3,202.2|3,208.1|
|Array map + reduce style loop (ops/s)|16,731,301|17,425,553|20,034,941|17,875,729|18,307,005|18,918,472|
|String build with join (MB/s)|3,417.7|3,438.9|3,480.5|3,589.9|3,498.6|3,581.6|
|Integer loop randomized (ops/s)|6,635,498|6,789,194|6,909,192|7,259,830|7,790,647|7,432,183|
If you're interested in the full charts and more details, you can check them out here: [Full Benchmark](https://www.repoflow.io/blog/python-3-9-to-3-14-performance-benchmarks). Let me know if you'd like me to run more benchmarks!
5 Answers
What OS did you run these tests on? It’s important to know to gauge performance accurately.
To really assess these benchmarks, I'd love to know how many times you repeated each test. Having some stats like standard deviation would also help clarify how significant these results are.
It seems like 3.11 was best for HTTP throughput, but I'm concerned about the json.loads regression. I think I’ll stick with 3.11 in production for now.
Interesting results! I noticed a drop in json ops from 3.9 to 3.14—can you elaborate on what might cause that?
Nice job on the benchmarks! Could you share the actual benchmark code? It'd also be helpful to annotate the charts with notes about what's better versus worse when it comes to the metrics you provided.
Definitely, ops per second makes it pretty clear—higher numbers are faster.
I ran this on macOS 25.0.0 with no background apps affecting performance.