Hi everybody!
I’m getting this error trying to use the gpt-40-search-preview model.
tools is not supported in this model. For a list of supported models, refer to https://platform.openai.com/docs/guides/function-calling#models-supporting-function-calling.
Any idea for the reason why?
Thanks in advance!
The reason why: the playground is broken.
It is sending an empty tools list, which is not tolerated on chat completions and not tolerated by the validation of the model.
Python:
response = client.chat.completions.create(
model="gpt-4o-search-preview",
messages=[
{
"role": "system",
"content": [
{
"type": "text",
"text": "You are a helpful AI, with integrated search results to improve your answering.\nUse search results as knowledge only, informing the facts and truthfulness of your response instead of merely reciting back. Produce no markdown web links. Follow the `user` message instructions."
}
]
},
{
"role": "user",
"content": [
{
"type": "text",
"text": "Provide a description of the hotel:\nRadisson Resort at the Port Cape Canaveral"
}
]
}
],
response_format={
"type": "text"
},
tools=[],
)
print(response.choices[0].message.content) to that…Drop the tools parameter entirely on the API and you will have success.
Except for no success in guiding this model by system message, because OpenAI re-prompts it. It will follow solely user wishes better.
The Radisson Resort at the Port, located at 8701 Astronaut Blvd in Cape Canaveral, Florida, is a 284-room hotel situated approximately 1.5 miles from Port Canaveral, making it a convenient choice for cruise passengers. (hotelguides.com)
Accommodations:
The hotel offers a variety of rooms and suites equipped with amenities such as flat-screen TVs, microwaves, mini-fridges, coffee makers, and complimentary Wi-Fi. Many rooms feature Sleep Number beds for personalized comfort. (oyster.com)
Dining:
Guests can dine at Flamingo’s Restaurant, which serves American cuisine with a Southern Florida flair in a casual setting overlooking the pool area. The poolside Tiki Bar offers refreshing drinks, allowing guests to enjoy the Florida sun while surrounded by swaying palms and lively music. (radisson-resort-at-the-port-cape-canaveral.booked.net)
Amenities:
The resort features a large heated outdoor pool with unique rock and waterfall features, a hot tub, and lighted tennis courts. Additional amenities include a fitness center, business center, and 15 meeting rooms, including a Grand Ballroom and pavilion, totaling 30,000 square feet of event space. (oyster.com, cvent.com)
Services:
The hotel offers cruise packages that include accommodations, parking, and shuttle service to Port Canaveral. An airport shuttle service is available for guests flying into Orlando International Airport. The property is pet-friendly, allowing pets up to 40 pounds with a fee of $35 per stay per pet. (portcanaveralcruiseport.com, hotelguides.com)
Location:
Situated on a busy road lined with hotels and restaurants, the Radisson Resort at the Port is less than five minutes by car from Port Canaveral and the Cocoa Beach Pier. Nearby attractions include the Kennedy Space Center Visitor Complex and SeaWorld® Orlando. (oyster.com, portcanaveralcruiseport.com)
The Radisson Resort at the Port combines comfort, convenience, and a range of amenities, making it an ideal choice for travelers visiting Cape Canaveral and the surrounding areas.
Similar question here but different error: Am testing out gpt-4o-mini-search-preview in my local to see what queries, snippets, url etc it fetches for a certain input query. The model appears in my dashboard but returns 404 Model not found via API.
This is as of yesterday (13 Dec, 2025).
When I complained about it with the OpenAI Support bot, it told me:
Access to preview models like gpt-4o-mini-search-preview is typically limited and may only be available to selected users or organizations during the preview phase. If you're seeing "404 Model not found" via the API: The model may not be generally available via API. Only accounts or organizations specifically included in the preview can access it.
But I see/read people using it and sharing analysis online - dated back as much as March 2025. So, the above response does not seem correct.
Mine is a test project with a minimum balance ($5), and I am the admin myself (of course).
Does it really require special permission? Or something to be switched/allowed in my dashboard?
I tested trying to use the model incorrectly as you might be doing. The Responses API endpoint returns 404 error as you report, because this is a search model for use on the Chat Completions API endpoint.
The reason it is only on Chat Completions is because Chat Completions doesn’t have tool calling for OpenAI’s internal hosted tools. On the Responses API, you need to use the web_search tool, and configure the tool with user location data.
This model will always answer as if your question was needing or wanting a web search results answer.
Thank you. I shall give it a try and see if it works 