Performance Very Slow compared with Github Copilot Chat

[jsa4000]

Describe the bug

Hi,

I have tested the same prompt using Github Copilot CLI and Github Copilot Chat with the same configuration, same MCPs, tools, agents, prompt, etc.. and while Github Copilot Chat takes 1-2 minutes, Github Copilot CLI takes 12-14 minutes. I have tested multiple times using different OS

I have extracted the logs and asked github copilot to check the timestamps to check the bottlenecks. I have attached the logs and the report generated in the logs.

As a summary, as far I can see so much time it's spend in flushing the events to the session. I don't know if the behavior in Github Copiloit Chat it's the same, what i know it is a lot fastest...

I am using gpt-5-mini model and several custom subagents that are delegated by the primary agent. These makes some calls to mcps and tools

This is the summary part I think it is the bottle neck.

- #1 — 227.882 s gap

  - Prev (line ~8790): 2026-01-14T06:30:59.394Z — "[DEBUG] Flushed 3 events to session ..."
  - Next (line ~8791): 2026-01-14T06:34:47.276Z — "[DEBUG] response (Request-ID ...):"
  - Interpretation: long wait between flushing events and receiving a response from a request to the AI model. Could be a model/API hang, queueing, or network stall.

- #2 — 105.220 s gap

  - Prev: 2026-01-14T06:35:54.504Z — "Flushed 4 events to session ..."
  - Next: 2026-01-14T06:37:39.724Z — "response (Request-ID ...):"
  - Interpretation: another long wait for model/API response.

Do you know what the problem could be? the use of subagents? the model response (even i am not experiencing the same problem in Github Copilot Chat )?

Thakns in advance
Regards

Affected version

0.0.380 Commit: 39ff7b7

Steps to reproduce the behavior

• Run copilot
• Select a custom agent
• Write a prompt

Expected behavior

No response

Additional context

process-1768372062780-8043.log

summary.md