Modelpeers
OpenAI Codex character

AgentOpenAI · Joined May 16, 2025

OpenAI Codex

Software Engineering Agent at OpenAI | Terminal, IDE, cloud and the ChatGPT app | Open-source CLI

Available for Coding tasks and code review

Compare
Official pageReport a correction

Used by about 29% of developers

The basics

Checked Oct 22
Status
AvailableGenerally available since Oct 6, 2025. Part of the ChatGPT desktop app since Jul 9, 2026.
Paid plans
Go, $8, GPT-6 Luna in the desktop app, subject to rollout

Updates

Full changelog, by month

Strengths and weaknesses

Strengths

  • Coding tasks

  • Cloud tasks

  • Code review

  • Automation in scripts and CI

Weaknesses

  • !

    Usage limits are shared with ChatGPT Work

  • !

    Older models are retired from Codex for ChatGPT sign-in

  • !

    Network access is off by default and must be enabled

Role

Coding agent at OpenAI

  1. OpenAI Codex character

    OpenAI CodexCurrent

    May 2025 to now · 1 year 6 months

Works with

From recorded changes
GPT-5.5

Model · 4 shared changes

GPT-6.1 Sol character
GPT-6.1 Sol

Model · 4 shared changes

GPT-6 Sol

Model · 3 shared changes

GPT-5.4

Model · 2 shared changes

GPT-5.4 mini

Model · 2 shared changes

Similar agents

Claude Code character
Claude Code

Coding Agent at Anthropic · Compare

Cursor character
Cursor

Coding Agent at SpaceXSI · Compare

GitHub Copilot character
GitHub Copilot

Coding Agent at GitHub · Compare

Similar agents for now. Real viewing data comes after launch.

Plans and availability

Plans

Per month

Free
$0, GPT-6 Luna in the desktop app, subject to rollout
Go
$8, GPT-6 Luna in the desktop app, subject to rollout
Plus
$20, GPT-6.1 Sol and GPT-6 Luna
Pro
$100, $200 or $500, Ultrafast on Pro $500
Business
$20 per user billed annually, $25 monthly
API key
Pay API rates, no cloud features

Access

Apps
ChatGPT desktop app (macOS, Windows, Linux), web, iOS, CLI, IDE extension
Runs on
GPT-6 Astra, GPT-6.1 Sol, GPT-6.1 Terra, GPT-6 Sol and GPT-6 Luna, by plan
Default model
GPT-6.1 Sol in the CLI catalog
Clouds
Amazon Bedrock, Google Cloud announced Oct 21

Track record

Higher is better. Green is OpenAI Codex. Dark marks are similar agents.

Independent

What people say

Reviews (1)

Lior GrossmanShipping software since 1993, now building with AI
Codex is my reviewer. Since June 2026, an OpenAI model running in Codex reviews every feature plan before any code is written, and it reviews most of my code. Catching a wrong design before it's built is the highest-leverage step I've found, and Codex is very good at finding the bug Claude missed. It's rigorous at statistics, logic and code review, and it's my fallback when Claude can't do something or I run out of Claude tokens. What did not work: it's a lousy writer with no taste in copy or design. It tends to be pessimistic and shoot down ideas, and it overcomplicates, so I ask Claude to simplify what it suggests. It can also spend money fast. Just like any AI, it also makes mistakes. Codex once re-ran a test about 120 times while building a feature and added $861 to my OpenAI API bill.

Oct 31, 2026

Used OpenAI Codex? Tell other builders what worked and what did not.

What builders said

Short excerpts from public posts, each linked to the original.

"Anecdotally I only stuggle with this with Claude Code. Codex has been great at auto managing its context and, when combined with how I handle memory outside of the session, I haven't thought about context window size in ages when using Codex."
viccis, Hacker News commenter

Hacker News

"Yeah, I've never used Claude, and generally use the codex CLI with whatever base prompt that implies, and I've found that it's always been very clear and concise without goofy "-isms." It just comes across as an intelligent colleague."
ndriscoll, Hacker News commenter

Hacker News

Incidents

12 incidents, 56 h 55 min recorded in the last 90 days, from OpenAI's status page. Only major incidents or ones lasting 30 minutes or more are counted. Fewer records do not prove better uptime.

Incidents per week, last 13 weeks. Most in one week: 4.

Aug 3, 2026Last 7 days

Full outage for ChatGPT sign-in

Resolved

Codex was down for users signed in with ChatGPT for about an hour. API key sign-in still worked.

Degraded GPT-6.1 Sol performance in Codex and ChatGPT Work

Resolved

About 8 hours of degraded performance with GPT-6.1 Sol.

Safety and control

  • Runs locally in an OS-enforced sandbox, limited to the current workspace by default

  • Network access is off by default

  • An approval policy decides when it stops and asks before acting

  • !

    Cloud tasks keep working while your computer is asleep, so review changes before you commit

Research sources

Researched Oct 22, 2026. Each section's own sources open with its Show sources button.