The Most Popular Coding Tools: Codex vs Claude Code
These days it's not just developers. Regular users are also turning to Codex and Claude Code for simple coding and agent tasks. But features aren't the only thing on your mind when you pick a coding tool. How much work you can actually hand off matters too.
In the US, on monthly billing, ChatGPT Plus and Claude Pro are each $20. ChatGPT Plus includes Codex, and Claude Pro lets you use Claude Code. (Both plans charge API usage separately, so I won't cover that here.)
So the two tools cost the same per month. But they differ in how usage is counted and in how many tokens were measured. The short version: going by public data as of November 2026, Claude Pro gives you a bit more room on tokens. You might find that surprising.
The common wisdom was that Claude's limits are tight and you hit them fast, right?
✦ Two Limits, Both Apply at Once
Hit either one and you stop
To start, both plans apply a 5-hour limit and a weekly limit at the same time. If you use up the 5-hour limit, you're cut off even with weekly limit left, and you have to wait for the reset.
Likewise, if you use up the weekly limit, you're cut off even with the 5-hour limit left, and you wait for the reset that comes around once a week.
✦ What Shares One Limit
How each subscription bundles its usage
Anyway, let's start by comparing how each one bundles its usage.
✦ ChatGPT Plus: Estimated Local Messages per 5 Hours
Estimates from OpenAI's pricing page
On ChatGPT Plus, Codex and ChatGPT Work share agent usage. Messages sent locally and cloud tasks come out of the same limit too.
Depending on the model, the estimated local messages you can send on Plus in 5 hours are 5 to 45 for GPT-6 Astra, 15 to 160 for GPT-6.1 Sol, 35 to 285 for GPT-6.1 Terra, 15 to 150 for GPT-6 Sol, and 350 to 3,000 for GPT-6 Luna.
Usage varies with the model, reasoning settings and task size, and cloud tasks can burn through more than local messages do.
✦ ChatGPT and Codex Usage: Today vs. December 2026
What changes on the Plus, Pro, Business plans
For now, regular ChatGPT chat usage is still separate from Codex's agent limit. That's why ChatGPT users use web chat and Codex separately.
But you should know that starting in December 2026, usage allowances for ChatGPT and Codex's agentic features on the Plus, Pro, Business plans are set to be merged on a token basis.
Instant chat is left out of the new shared limit, but still, it feels like the era of using web chat as freely as before is passing. Maybe it's because OpenAI's fundraising results have been weak lately, but it's starting to tighten usage limits bit by bit.
Claude Pro has always had Claude on web, mobile and desktop and Claude Code split the same subscription usage. How much each message uses depends on message length, attachments, conversation length and the model.
So Claude Code users were wary of using web chat on a paid Claude plan. Use web chat and your Claude Code usage shrinks along with it.
That's where the sense came from that ChatGPT has roomier usage than Claude, but in the end ChatGPT is adopting this kind of usage model too.
Claude Pro, Which Raised Its Limits
✦ May 2026: Claude Code Limits Went Up
Based on the Claude Code 5-hour limit
Claude Code's usage jumped a lot in May 2026. On May 6, the 5-hour limit on Pro, Max, Team and other major plans doubled, and the peak-hour deduction that applied to Pro and Max went away.
Pro subscribers can now get more Claude Code work done in the same 5-hour session than before. You really feel it. Back then, tokens melted like ice cream in summer.
✦ Two Ways to Measure Usage
Monthly estimate and weekly median
Neither ChatGPT Plus nor Claude Pro publishes its total token count, but someone ran an experiment to estimate it.
They measured how fast the usage limit shrinks for each of input, cache write, cache read and output tokens, then used that to estimate monthly usage.
The result: ChatGPT Plus with GPT-6.1 Sol came out to about 1 billion tokens, and Claude Pro with Opus 5.5 to about 2.9 billion. So for the same $20, the math says Claude Pro can process about 2.9 times as many tokens.
Priced at each model's list API rates, ChatGPT Plus works out to about $211 and Claude Pro to about $1,178. With Claude Pro, that's about 5.6 times more usage by API pricing.
The token gap was 2.9x, but it widened because the API token prices of Opus 5.5 and GPT-6.1 Sol differ.
Account-level weekly tracking shows a similar gap. In Wokey's November update, Claude Pro had a weekly median of 24.2 million uncached tokens across 79 accounts, with 1.7 million output tokens.
Codex on ChatGPT Plus had 6.8 million uncached tokens and about 390,000 output tokens across 12 accounts. Excluding reused cache, Claude Pro was about 3.6 times higher.
The sample sizes and observation periods differ, but even allowing for that, Claude Pro looks roomier on token usage.
✦ Image Generation: Where They Part Ways
Codex vs. Claude Code
Of course, Codex's strength is that you can use ChatGPT, the best at image generation, together with a coding agent under one subscription plan. Claude Code, on the other hand, has no image generation of its own.
(Sure, people on YouTube say Claude makes amazing slides, but I'm not that impressed...) ChatGPT's free image generation limits feel like they're getting tighter lately, so I think Codex will still be appealing if you generate a lot of images.
Same $20, but Claude Pro Gives More Coding Usage
Looking at the limits that are public and the unofficial measurements, the plan with roomier usage for coding agents on a $20-a-month subscription looks to be Claude Pro.
And if OpenAI merges web chat into the Codex limit going forward, I wonder if Codex's appeal won't drop even further. If it keeps going like this, I do have some doubts about whether OpenAI can turn a profit, post a surprise earnings beat and pull off an IPO.
Let's keep watching together how ChatGPT and Claude plans change from here.
That's all for this post. See you next time.
Related posts
- Do Paid AI Assistants Beat Free Ones? Test Results
- GPT-6 Astra and Claude Fable 5.1 Failed a Paper Review
- Claude Motion Turns Reports into Animation (Team/Enterprise)
🌐 Korean version: 같은 월 20달러인데, 사용량은 어느 쪽이 넉넉할까? OpenAI Codex vs Claude Code 비교!
No comments:
No comments yet. Want to leave the first one?
Post a Comment