Skip to content

Why have agents (or GPT 4.1) slowed down so much? #1249

@sibblegp

Description

@sibblegp

Describe the bug

I have not changed anything in my code but all of my agent calls are significantly slower. I'm using GPT-4.1. One has gone from under 4 seconds to over 11 seconds. All of that extra time is spent in the responses API.

We can easily see the results on our client apps. It takes much longer to process and the text is rendered slower.

This trace was from 3 days ago:

Image

This one is from today, notice the difference in time in the POST /v1/responses section and that the tool call on my end took the same amount of time.

Image

All of my calls are like this taking significantly longer to process.

This is a major problem and has been going on for 2 days straight. OpenAI's status page says nothing but something huge changed. This is kind of urgent since the response time more than doubled. This is a hosted server on GCP and locally. It's the AI model that is significantly slower.

Debug information

  • Agents SDK version: v0.2.3
  • Python version: 3.13

Repro steps

Use an agent with a tool call that processes JSON to format a response. It's notably slow.

Expected behavior

Fast responses.

Metadata

Metadata

Assignees

No one assigned

    Labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions