Serve a ML model on a single machine with Flask + Gunicorn vs. FastAPI + Uvicorn

# server.py
import fastapi
import flask
import spacy

# load NER model
ner = spacy.load("en_core_web_sm")

# flask web server
flask_app = flask.Flask(__name__)

@flask_app.get("/")
def ner_flask():
    text = flask.request.json.get("text")
    entities = ner(text)
    return {ent.text: ent.label_ for ent in entities.ents}

# fastapi web server
fastapi_app = fastapi.FastAPI()

@fastapi_app.get("/")
async def ner_fastapi(request: fastapi.Request):
    text = (await request.json()).get("text")
    entities = ner(text)
    return {ent.text: ent.label_ for ent in entities.ents}

# client.py
import asyncio
import time

from aiohttp import ClientSession

api_url = "http://127.0.0.1:5000/"

text = """Paris[a] is the capital and most populous city of France. With an official estimated population of 2,102,650 residents
as of 1 January 2023[2] in an area of more than 105 km2 (41 sq mi),[5] Paris is the fourth-most populated city in the European
Union and the 30th most densely populated city in the world in 2022.[6] Since the 17th century, Paris has been one of the world'
major centres of finance, diplomacy, commerce, culture, fashion, and gastronomy. For its leading role in the arts and sciences,
as well as its early and extensive system of street lighting, in the 19th century, it became known as the City of Light.[7]
The City of Paris is the centre of the Île-de-France region, or Paris Region, with an officia estimated population of 12,271,794
inhabitants on 1 January 2023, or about 19% of the population of France.
"""

async def fetch(session: ClientSession, i_request: int):
    # fetch NER result for a request
    async with session.get(api_url, json={"i_request": i_request, "text": text}) as response:
        result = await response.json()
    return result

async def main(i_trial: int):
    # send 1000 simultaneous requests to the server
    num_requests = 1000
    async with ClientSession() as session:
        tasks = []
        for i_request in range(num_requests):
            tasks.append(fetch(session, i_trial*num_requests + i_request))
        await asyncio.gather(*tasks)

num_trials = 5
times = []
for i_trial in range(num_trials):
    start = time.perf_counter()
    asyncio.run(main(i_trial))
    end = time.perf_counter()
    times.append(end - start)

print(f"Avg Time: {sum(times)/len(times):.2f}")

# server.py
@flask_app.get("/")
def ner_flask():
    time.sleep(0.1) # do some IO operations
    text = flask.request.json.get("text")
    entities = ner(text)
    return {ent.text: ent.label_ for ent in entities.ents}

@fastapi_app.get("/")
async def ner_fastapi(request: fastapi.Request):
    await asyncio.sleep(0.1) # do some IO operations
    text = (await request.json()).get("text")
    entities = ner(text)
    return {ent.text: ent.label_ for ent in entities.ents}