Running Agents with GPT-5 models causes arbritary long wait before ending the streaming after results have finished sending.

[JoelCCodes]

Please read this first

• Have you read the docs? Agents SDK docs
  Yes
• Have you searched for related issues? Others may have faced similar issues.
  Yes

Describe the bug

It seems that when running with GPT-5 models and hosted web_search, it creates an arbritary long wait after the response has finished. Sometimes up to 2-3 mines.

Debug information

• Agents SDK version: V0.0.14

Repro steps

Running a streamed Agent run using GPT-5 or GPT-5 mini and attach the web_search tool.

Expected behavior

I expect the stream to end right away instead it waits for a while before closing out.

[bakikucukcakiroglu]

I think even without web_search, it takes extremely long with gpt-5 even for simplest queries. I think this is related to Responses API. Do you use Responses API or Completions API?

Please see OpenAI Developer Community: GPT-5 + Responses API is extremely slow

[JoelCCodes]

@bakikucukcakiroglu I am using Responses API for my Agents, may be related.

[Doc-Code]

I have this problem too

[JoelCCodes]

Hopefully this gets some visibility soon!

[eddymendez1]

I've also run into this issue

[seratch]

We appreciate this feedback, and we're aware of the issue. We will look into viable solutions, but for now, please continue using gpt-4.1 as the default model for most use cases.

see also: openai/openai-agents-python#1397 (comment)

[rednikisfun]

I think even without web_search, it takes extremely long with gpt-5 even for simplest queries. I think this is related to Responses API. Do you use Responses API or Completions API?

Please see OpenAI Developer Community: GPT-5 + Responses API is extremely slow

I've used Chat Completions with GPT-5, and in comparison with GPT-4.1, the responses take much longer. I think it's an issue with the model itself. Didn't get the actual metrics yet though.