openai / openai/codex

ChatGPT / Codex Windows desktop client consumes excessive network traffic (~31 GB in 58 hours)

Open
#40,513 1 comment 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

app bug connectivity windows-os
Dominant language
Rust
Stars
125k
Forks
19.4k
PR merge metrics
PR metrics pending

Description

What version of the Codex App are you using (From “About Codex” dialog)?

Powered by Codex & OWL 版本 26.818.41509

What subscription do you have?

ChatGPT Plus

What platform is your computer?

Windows 11 (64-bit)

What issue are you seeing?

Summary

The ChatGPT / Codex Windows desktop client is generating an abnormally large amount of network traffic.

Windows network usage statistics show that the ChatGPT/Codex client downloaded approximately 31 GB of data within about 58 hours, despite normal usage that should not reasonably generate this amount of traffic.

Additional Windows network statistics also show that ChatGPT consumed 16.73 GB within the most recent 24-hour period.

The traffic is overwhelmingly download traffic rather than upload traffic.

This appears to indicate abnormal background network activity, repeated resource downloading, synchronization, reconnect loops, or another network-related regression in the Windows desktop client.

Environment
Platform: Windows
Application: ChatGPT / OpenAI Codex Desktop
Package name:

OpenAI.Codex_26.818.2441.0_x64__2p2nqsd0c76g0

App ID: 10056
Network connection: Wi-Fi
Network adapter: Intel(R) Wi-Fi 7 BE201 320MHz
Connection type: WLAN
Observed Network Usage
Record 1

Time period:

2026-08-21 23:34 → 2026-08-24 09:46

Network usage attributed to the OpenAI Codex application:

Download / received: 30.993 GB
Upload / sent: 0.518 GB

The download-to-upload ratio is approximately 60:1.

Approximately 31 GB was downloaded within about 58 hours.

This corresponds to roughly 12–13 GB of downloaded data per day.

Record 2

Windows Settings → Network & Internet → Advanced network settings → Data usage

During the most recent 24-hour period:

Total WLAN traffic: 22.11 GB
ChatGPT: 16.73 GB
Windows Update/System: 1.22 GB
WeChat: 519 MB
Other applications: substantially lower usage

ChatGPT alone accounted for approximately 76% of all network traffic during this 24-hour period.

Record 3

Hourly network monitoring also shows unusually large download spikes.

Example:

2026-08-24 11:00

Download: 635 MB
Upload: 14.8 MB

This again shows that the abnormal traffic is predominantly inbound/download traffic.

Actual Behavior

The ChatGPT/Codex Windows client consumes several gigabytes of network traffic per day.

The amount of downloaded data is far beyond what would normally be expected from text conversations, code generation, model responses, or ordinary application updates.

The network activity appears to continue beyond normal interactive usage and can result in tens of gigabytes of data consumption within only a few days.

Expected Behavior

The ChatGPT/Codex desktop client should only download data necessary for:

user conversations;
model responses;
required application resources;
intentional file transfers;
application updates.

Normal usage should not result in 10+ GB of background downloads per day.

When the application is idle or no active task is running, background network traffic should remain minimal.

Steps to Reproduce
Install/run the current ChatGPT/Codex Windows desktop application.
Connect through Wi-Fi.
Use ChatGPT/Codex normally.
Leave the desktop client running for an extended period.

Open:

Windows Settings → Network & Internet → Advanced network settings → Data usage

Check ChatGPT network consumption.
Compare it with operating-system-level network usage records.

In my case, the client downloaded approximately:

30.993 GB between 2026-08-21 23:34 and 2026-08-24 09:46.

Frequency

The issue is reproducible over multiple hours/days.

Network monitoring shows repeated large download activity rather than a single isolated application update.

Impact

This issue has a significant practical impact:

consumes tens of gigabytes of network data;
may exceed metered or corporate network quotas;
can cause unexpected mobile hotspot/data charges;
consumes company bandwidth;
makes it difficult to safely leave the ChatGPT desktop application running;
may indicate an underlying resource-fetching, caching, synchronization, WebSocket reconnect, session recovery, or update-loop problem.
Suspected Areas

Possible causes may include:

repeated downloading of frontend/static resources;
ineffective or broken caching;
repeated session/history synchronization;
connection or WebSocket reconnect loop;
repeated recovery of Codex task/session state;
application updater repeatedly downloading resources;
Electron/WebView resource reload loop;
background process continuing network requests when no foreground operation requires them.

These are only hypotheses. The operating-system statistics clearly identify the ChatGPT/Codex application as the source of the abnormal network traffic.

Evidence

I can provide screenshots showing:

Codex application network record:
30.993 GB received
0.518 GB sent
2026-08-21 23:34 → 2026-08-24 09:46
Windows 24-hour data usage:
ChatGPT: 16.73 GB
Total WLAN: 22.11 GB
Hourly network usage:
11:00 download: 635 MB
11:00 upload: 14.8 MB

Please let me know if additional diagnostic information would be useful, such as application logs, process-level TCP connection records, destination domains/IPs, or Windows network traces.

Image
What steps can reproduce the bug?

no

What is the expected behavior?

No response

Additional information

No response

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by examining the Windows network-usage records and collecting the application logs, process-level TCP connection records, destination domains or IPs, and Windows network traces offered in the report. Compare activity while the client is idle with active usage; done means identifying the source of the repeated downloads and confirming that idle traffic is minimal.

Written by the indexing model from the issue text.

Assessment

Tech stack
rust
Domain
desktop, networking
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Active
Clarity
Needs clarification
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.