#14962 replaced json.dumps with pydantic v2 TypeAdapter.serialize_json(). This is the same, but for body parsing - replace json.loads with TypeAdapter.validate_json().
See also #13949

Now this gives only a moderate speedup, nothing huge. The larger the request body, the larger the speedup.
Currently optimization can be applied only to POST handler with just one body field.

fastapi
locust
py-spy

from http import HTTPStatus

from fastapi import FastAPI, Response
from pydantic import BaseModel

app = FastAPI()

class SimpleModel(BaseModel):
    a: str


@app.post("/data")
async def create_item(data: SimpleModel):
    return Response(status_code=HTTPStatus.NO_CONTENT)

if __name__ == "__main__":
    import uvicorn
    uvicorn.run(app, host="0.0.0.0", port=8000)

import json

from locust import FastHttpUser, TaskSet, task


class FastAPITask(TaskSet):
    data = json.dumps({"a": "abc" * 10000})
    headers = {
        "Content-Type": "application/json",
    }

    @task
    def push_data(self):
        with self.user.rest(
            "POST",
            "data",
            data=self.data,
            headers=self.headers,
            name="/data",
        ) as response:
            response.js


class FastAPIUser(FastHttpUser):
    host = "http://localhost:8000/"
    tasks = [FastAPITask]

locust --web-host=0.0.0.0 --processes 4 FastAPIUser

record --duration 600 --output flamegraph.svg -- python -m uvicorn app:app --host 0.0.0.0 --port 8000 --log-level=warning

Before - 2228RPS, json.loads took 9.8% and type_adapter.validate_python took 1% = 11% combined
fastapi_with_json_loads.tar.gz

After - 2570RPS, type_adapter.validate_json took 7.17%
fastapi_with_typeadapter_json.tar.gz

Caveats - if route handle contains multiple fields with Body() annotation, json will be parsed for each field. Changing this requires substantial rewrite of body parsing, e.g. merging all body fields into one pydantic class and using it's model_validate_json() method for body parsing. I'm not ready for this type of change, so this can be considered as proof-of-concept.

Bump pydantic minimal version to 2.10.0 because of this issue with validate_json appears on versions below pydantic-core 2.24.1