OpenAI Codex CLI
Close to the top coding-agent score at a fraction of the cost per task; needs ChatGPT Plus or an API key.
Should you use it?
Pick it if
ChatGPT Plus subscribers who want a terminal coding agent at no extra cost.
Best for: ChatGPT Plus subscribers who want a terminal coding agent at no extra cost (Coding agents Top 20)
Skip it if
You don't pay for ChatGPT and want a free terminal agent; Copilot's CLI is on its free plan.
You give up: Default model unnamed; Plus limits vary widely and personal plans train by default
The short version
- What this is
- OpenAI's open-source coding agent for the terminal. Sign in with a ChatGPT plan or use an API key; latest stable release 0.162.1 (9 Oct 2026).
- Why it matters
- If you already pay for Plus, it costs nothing extra; if not, it isn't the strongest pick on current evidence.
- Open source, but real use needs ChatGPT Plus at $20/month or an API key s5 s8.
- Scores 70 for coding on our checklist; Artificial Analysis rates it 63 with GPT-6.1 Sol (xhigh) at $1.04 a task s11.
- Turn off training in ChatGPT data controls and in the Codex environments setting s6.
Good at
- Open source under Apache-2.0, with about 129k GitHub stars s8.
- Scores 63 on Artificial Analysis's Coding Agent Index v1.5 with GPT-6.1 Sol (xhigh) at $1.04 a task, against $13.04 for Claude Code at 66 with Opus 5.5 s11.
- DevDay 2026 added voice control and an /agents view to run several tasks at once s9.
Our take
The CLI itself is open source s8, but in practice it is part of a ChatGPT subscription: real use needs Plus at $20/month or an API key s5. The latest stable release, 0.162.1, came out on 9 Oct 2026 s10, and DevDay added voice control and a view for running several tasks at once s9.
On Artificial Analysis's Coding Agent Index v1.5, in use since September 2026 s13, Codex with GPT-6.1 Sol (xhigh) scores 63 for $1.04 a task, against 66 for Claude Code with Opus 5.5 (max) at $13.04 s11. Artificial Analysis lists the agent as Codex without naming the surface. Snorkel's Terminal-Bench 2.1 board, which does name Codex CLI, puts it first with GPT-6 Astra at 87.4% (3 Sep 2026), but that board has only 18 entries s12. The score is 70, not higher, because of the $20 plan, the wide limit estimates and training on by default s5 s6. OpenAI recommends GPT-6.1 Sol, fourth of 142 on arena.ai WebDev, but doesn't say what the CLI uses if you don't choose, so set it yourself s3 s4.
Score
A score from 0 to 100 per job, against everything available today. Overall 70 across 1 job, the average of the job scores.
New to this? What the scores and labels mean
- 0 to 100
- Our score for one job, like writing or coding, compared with everything else out today. 100 is the best available today. For AI models we build it from public data, such as Arena's blind votes and Epoch AI's independent tests; apps and tools are scored on our own checklist. The same tool can score differently for different jobs, and where two likely ranges overlap, we call it a tie.
- Our verdict
- Use it: a top choice for most people. Worth trying: good, and worth a go. Depends: good for some people or one kind of job. Wait: too early to tell, or not out for everyone yet. Skip: something else does the job better.
- First look
- Just released. We say what it is and whether to bother, and give it a score once public results exist.
- No score yet
- None of the public sources we use covers it yet, or the only evidence is from a source whose numbers we can't use, so we describe it in words instead.
- Best buy
- Nothing cheaper scores as well for that job.
- Best free
- The best score for that job that costs nothing, through a free plan or by running on your own computer.
- Pay as you go
- No monthly plan: you pay for each use, usually through a developer account.
- Checked
- The day we last confirmed the score and the price against the sources. Prices change often, so check the date.
- AI model
- The engine underneath, like GPT or Claude. You use it through an app or, if you code, directly.
- Open source
- Free to download and run on your own computer, so nothing leaves your machine.
Our checklist score, going by an independent test, a public leaderboard and the maker’s pages.
Our checklist
2 sources
Sources and notes
Independent testPublic leaderboardMaker’s pagesChecklist 7/10: does the job 2 (Artificial Analysis Coding Agent Index v1.5, in use since September 2026 s13, scores Codex 63 with GPT-6.1 Sol (xhigh) at $1.04 a task, against 68 for the best Claude Code row s11; Snorkel's Terminal-Bench 2.1 board lists Codex CLI with GPT-6 Astra first at 87.4%, submitted 3 Sep 2026, but has 18 entries, so it is a note only s12); underlying quality 2 (GPT-6.1 Sol, which OpenAI recommends, is arena.ai WebDev rank 4 of 142, gpt-6.1-sol-max s3 s4); value 1 (needs ChatGPT Plus at $20/month or an API key s5); privacy 1 (personal plans train unless you opt out s6); limits 1 (Plus allows an estimated 15 to 160 GPT-6.1 Sol messages per five hours, and weekly caps may apply s5).
Alternatives
Better or cheaper for coding, from our scores, then the ones we'd also look at. The same kind of product comes first.
- OpenCodeBetterCoding agent80FreeScores 80 in coding, against 70 here; free. Side by side
- Claude CodeCoding agent70$20 a monthScores a little higher on Artificial Analysis's test (68 against 63) but costs far more per task; also $20/month. Side by side
- GitHub Copilot (including Copilot CLI and cloud agent)Coding agentFirst look: Worth trying$10 a monthIts CLI comes with the free plan; Pro is $10/month. Side by side
- OpenAI Codex CloudCoding agentFirst look: Worth trying$20 a monthThe same Codex agent running in OpenAI's cloud while your computer is off. Side by side
Value for money
Where it sits in coding: higher is a better score, further left is cheaper. Green dots are the Best buys.
Evidence
Everything behind the call and the scores, folded so the page stays short. Spotted a wrong price, score or link? Report a mistake.
How we scored this
The coding score comes from our own checklist of the product.
Independent testPublic leaderboardMaker’s pages
Reviews are drafted with help from AI models, including Anthropic's Claude. Model scores come only from outside leaderboards and independent evaluations, by a written rule.
No one paid for this review or gave us free access, and we earn nothing if you sign up. How we score
Sources 13
- s1Artificial Analysis: Cursor Composer 2.5 on the Coding Agent Index (20 May 2026)
- s2Artificial Analysis: Coding agents leaderboard
- s3Codex docs: Models
- s4arena.ai Code Arena: WebDev leaderboard
- s5Codex docs: Pricing
- s6OpenAI Help: How your data is used to improve model performance
- s7arena.ai Agent leaderboard
- s8openai/codex on GitHub
- s9OpenAI: DevDay 2026 recap
- s10openai/codex releases
- s11Artificial Analysis: Claude Code vs Codex, Coding Agent Index v1.5
- s12Snorkel AI: Terminal-Bench 2.1 leaderboard (updated 1 Oct 2026)
- s13Artificial Analysis: Coding Agent Index v1.5 methodology
Something here out of date or wrong? Report a mistake and we will check it against the source.
History 3 changes
Every change to a rating, the price or our call, with the reason. Scores are against what exists on the day, so they drift down as better things ship. Older entries come from before we moved to scores.
- First look. Independent agent results are older than 90 days or test older models, so not rated yet.
- Added Artificial Analysis Coding Agent Index v1.5 score: 63 with GPT-6.1 Sol (xhigh). No per-result date and the main index hides values, so still a First look.
- New: First look to ScoredNew: Coding ratedRated for coding by the checklist. Artificial Analysis's methodology page shows v1.5 has been in use since September 2026, so its scores fall inside the 90-day window. Agents not rated yet.
Use it well 4 guides on this site
Pages on this site that help with the jobs above.