I have shipped production APIs in Flask, Django REST Framework, and FastAPI. FastAPI is the correct choice for new Python API development, and the gap is not close.
Type-hinted request/response validation via Pydantic, automatic OpenAPI documentation generated from your code, async support built-in, and dependency injection that is clean and testable — in a framework that benchmarks within 10% of Go and Node.js for IO-bound workloads.
The Sabivox ML inference API is FastAPI. Async support handles concurrent inference requests. Pydantic models catch malformed input before it reaches the model. Auto-generated documentation is what the ML team uses to understand the contract.
For pure API development — especially ML-serving and async workloads — FastAPI is the answer. Start there.
— Dick Bassey | DevDick | 2024