The model card for o3-pro currently says it supports Batch APIs. https://platform.openai.com/docs/models/o3-pro.
When trying to use the o3-pro in the batch API, however, I get the following response:
Batch(id='batch_684a4bc5b0ac81909a63f5b8e20c4809', completion_window='24h', created_at=1749699525, endpoint='/v1/responses', input_file_id='file-TYJs2sKihpccEu5t3CTS9W', object='batch', status='failed', cancelled_at=None, cancelling_at=None, completed_at=None, error_file_id=None, errors=Errors(data=[BatchError(code='model_not_found', line=1, message="The provided model 'o3-pro' is not supported by the Batch API.", param='body.model'), BatchError(code='model_not_found', line=2, message="The provided model 'o3-pro' is not supported by the Batch API.", param='body.model')], object='list'), expired_at=None, expires_at=1749785925, failed_at=1749699527, finalizing_at=None, in_progress_at=None, metadata={'description': 'nightly eval job'}, output_file_id=None, request_counts=BatchRequestCounts(completed=0, failed=0, total=0))
Here is the code I am using to query the Batch API:
from openai import OpenAI
import time
client = OpenAI()
batch_input_file = client.files.create(
file=open("batchinput.jsonl", "rb"),
purpose="batch"
)
batch =client.batches.create(
input_file_id=batch_input_file.id,
endpoint="/v1/responses",
completion_window="24h",
metadata={
"description": "nightly eval job"
}
)
print("ID: ", batch.id)
while True:
batch = client.batches.retrieve(batch.id)
print(batch)
if batch.status == "completed" or batch.status == "failed":
break
time.sleep(5)
if batch.output_file_id:
print("File Output:")
output_file = client.files.content(batch.output_file_id)
print(output_file.text)
if batch.error_file_id:
print("File Error:")
error_file = client.files.content(batch.error_file_id)
print(error_file.text)
And here is the jsonl file:
{"custom_id": "request-1", "method": "POST", "url": "/v1/responses", "body": {"model": "o3-pro", "input": [{"role": "system", "content": "You are a helpful assistant."},{"role": "user", "content": "Hello world!"}],"max_output_tokens": 10000}}
{"custom_id": "request-2", "method": "POST", "url": "/v1/responses", "body": {"model": "o3-pro", "input": [{"role": "system", "content": "You are an unhelpful assistant."},{"role": "user", "content": "Hello world!"}],"max_output_tokens": 10000}}
I have verified this code works for other models like GPT-4o or o1. Using o3-pro-2025-06-10 produces the same behavior.
