Consistently getting 503 Error when using Gemini-2.5-Pro model using Google GenAI SDK

[Madhusree_Rana]

I had written the following code:

import os

import time

from dotenv import load_dotenv

from google import genai

from google.genai import types

from prompts import *

from schemas import GroupBenefit

def extract_parameters(

client,

model,

contents,

config

):

print(f"\nAttempting extraction using model: {model}")

response = client.models.generate_content(

model=model,

contents=contents,

config=config

)

return response

def print_result(model: str, response: dict, start_time: float):

print(f"\nExtraction successful using model: {model}!")

print(response)

print(“\nTotal time taken: %s seconds” % round((time.time() - start_time), 2))

if _name_ == “_main_”:

start_time = time.time()



load_dotenv()

MAIN_MODEL = os.getenv('MAIN_MODEL')

MINI_MODEL = os.getenv('MINI_MODEL')

TEMPERATURE = float(os.getenv('TEMPERATURE'))



client = genai.Client()



input_file_path = "OutputFiles/Renewal 2025_extracted.txt"

try:

file = client.files.upload(

file=input_file_path,

config={‘mime_type’: ‘text/plain’}

    )



    contents = \[

        get_gb_prompt_for_excel(naic="23"),

file

    \]



    config = types.GenerateContentConfig(

system_instruction=get_system_prompt(),

temperature=TEMPERATURE,

response_mime_type=‘application/json’,

response_schema=GroupBenefit

    )



    model = MAIN_MODEL

    response = extract_parameters(

client=client,

model=model,

contents=contents,

config=config

    )

if hasattr(response, “parsed”) and response.parsed:

        print_result(model, response.parsed.model_dump_json(indent=2), start_time)

except Exception as e:

    err = str(e).lower()

if any(sub in err for sub in [“503”, “unavailable”, “overloaded”]):

print(f"Model ({model}) overloaded. Retrying with another model…")

try:

            model = MINI_MODEL

            response = extract_parameters(

client=client,

model=model,

contents=contents,

config=config

            )

if hasattr(response, “parsed”) and response.parsed:

                print_result(model, response.parsed.model_dump_json(indent=2), start_time)

except Exception as e:

print(“\nExtraction failed after all retries!”)

print(str(e))

raise

else:

print(err)

raise

====================================================================

The main and mini models are as follows:

MAIN_MODEL=gemini-2.5-pro
MINI_MODEL=gemini-2.5-flash

Whenever I am running the code, I am almost always getting the following error for “gemini-2.5-pro”, and sometimes for “gemini-2.5-flash”:

503 UNAVAILABLE. {‘error’: {‘code’: 503, ‘message’: ‘The model is overloaded. Please try again later.’, ‘status’: ‘UNAVAILABLE’}}

I am using the free tier of the API.

Can you please resolve this issue at the earliest?

[Cyberknight]

i am having the same issue on my end

[Gene13]

same here over the last few days

[James_McCabe]

Yes same issue here today using either 2.5 pro or flash. Model overload. Changed my location env. variables from us-central1 to us-east4. Worked for a bit and then same errors.

[Pranav_Singhania]

We are observing the same behaviour across our apps, and Google’s page does not acknowledge any degradation of service surprisingly

[Rob_O]

Exactly the same here. New customer, not on the free tier. Heres the situation yesterday. Currently I couldn’t go anywhere near this for production. Simpler prompts get through 2/3 times, the more involved ones take 5-10 refreshes before it finally gets through. All 503’s.