There's a project called django-bolt making the rounds. The pitch: replace Django's WSGI/ASGI server with an Actix Web layer written in Rust, via PyO3. Claims 60k-100k requests per second on pure endpoints, ~13k on ORM reads.
I went down the rabbit hole. Here's my honest take.
What it actually is.
Not a Django replacement. Not a rewrite. An add-on that swaps the HTTP layer while leaving everything else intact - ORM, Admin, models, middleware, Celery, all of it stays. You install it, add it to INSTALLED_APPS, run python manage.py runbolt instead of gunicorn. That's the interface.
The migration path is endpoint-by-endpoint. Serializers move from DRF to msgspec.Struct. Views move from ViewSets to @api.get decorators. Auth needs wiring. On a codebase with 20-30 endpoints, the estimate is 3-5 days.
The numbers are real. The caveats matter.
60k-100k RPS on pure endpoints is impressive. For context, a well-tuned gunicorn setup on comparable hardware typically caps out significantly lower. Rust's concurrency model is genuinely faster at the HTTP layer.
The 13k RPS on ORM reads is more interesting because that's where real Django apps spend most of their time. Still faster than most Python setups. Still bounded by database I/O.
The caveat: this is a one-person project with ~2k GitHub stars. It's under active development. "Not production-ready unless raw performance is a hard requirement" is the honest summary.
When it would and wouldn't make sense.
Worth considering if: you have a Django app where HTTP throughput is a genuine bottleneck, you've already done the obvious Python-level optimizations, and you have the engineering bandwidth to manage a dependency on an early-stage project.
Not worth it if: your bottleneck is the database (it usually is), you need stability guarantees, or the 3-5 day migration cost isn't justified by the performance gain you'd actually see in production.
One thing I went looking for and didn't find: a cookiecutter-django-bolt. Nothing on GitHub. If this project matures, that's the missing piece - a batteries-included starting point the way cookiecutter-django is for standard setups.
The broader signal.
The real story here isn't django-bolt specifically. It's that Rust is showing up at the edges of the Python ecosystem with increasing regularity - not replacing Python, but taking over the performance-critical layers where Python was always the bottleneck.
Pydantic v2 did this. Polars did this. Ruff did this. Django-bolt is attempting the same at the HTTP layer.
The pattern: keep Python's ergonomics and ecosystem, replace the hot path with Rust. It's a good pattern. Watch where it goes.
TL;DR
- django-bolt replaces Django's HTTP server with Rust/Actix Web via PyO3. ORM, Admin, everything else unchanged.
- Claims 60-100k RPS on pure endpoints, ~13k on ORM reads. Numbers are real; project is early stage.
- Worth it only if HTTP throughput is your actual bottleneck - which it usually isn't.
- The bigger signal: Rust eating Python's performance-critical layers is a pattern, not a one-off.