AI coding plan usage limits compared (October 2026)

Published 2026-10-20 · Limits checked against official pages (Oct 18–20, 2026) · 9 min read · Not affiliated with any vendor

TL;DR

Claude Code: rolling 5-hour session plus weekly limits, shared with the Claude apps. Only multipliers are published (Max = 5x or 20x Pro).^[1]
Cursor: one weekly pool for Cursor models, sized 1x / 4x / 15x / 50x. Claude, GPT and Gemini are outside the pool and bill at API price.^[3]^[4]
Codex: ChatGPT Plus has published message ranges per 5 hours; Pro currently has no 5-hour limit, but weekly limits may apply.^[5]
Copilot, OpenCode Go and Zed meter in dollars, so you can actually predict when you run out.^[6]^[9]^[10]
Nobody publishes how many tokens a Claude, Cursor or Devin “1x” is. Estimate your real monthly cost →

Why limits suddenly matter

After Cursor's October 12 repricing, a lot of developers are comparing plans by what they can actually do per week, not by sticker price. You can see it on X. Posts matching our query for usage-limit terms (“usage limit”, “weekly limit”, “5-hour limit”, “rate limited”) next to Claude Code, Cursor or Codex ran at 166–170 a day on October 14–16 (UTC). On October 17–19 they rose to 207–225 a day. That is a keyword count, not a sentiment analysis, but most of those posts are people hitting a cap and asking where the cap is.

The problem: most vendors describe limits in relative terms, so the comparison has to be made unit by unit. Here is what each one officially publishes.

Tool	Paid plans	What's metered	Reset windows	Absolute size published?	When you hit it
Claude Code^[1]	Pro $20 · Max 5x $100 · Max 20x $200	Relative usage (Max = 5x / 20x Pro per session)	Rolling 5-hour session + weekly	Multipliers only	Usage credits at API rates, or API pay-as-you-go
Cursor (new plans)^[3]	Plus $8 · Premium $30 · Super $100 · Ultra $200	Pool for Cursor models (4x / 15x / 50x Plus), shared with Grok	Weekly	Multipliers only	Prepaid top-ups; Claude/GPT/Gemini always at API list price
OpenAI Codex^[5]	Plus $20 · Pro $100/$200/$500	Local messages per 5 hours (by model)	Plus: 5-hour + weekly · Pro: no 5-hour limit, weekly may apply	Ranges, e.g. GPT-6.1 Sol 15–160 per 5h on Plus	ChatGPT credits, or an API key
GitHub Copilot^[6]	Pro $10 · Pro+ $39 · Max $100	AI credits ($0.01 each): $15 / $70 / $200 per month	Monthly	Yes, in dollars	Dollar budget at $0.01 per credit
Devin Desktop (Windsurf)^[7]	Pro $20 · Max $200	Quotas	Daily + weekly refresh	No numbers	Extra usage at API pricing
Kiro^[8]	Pro $20 · Pro+ $40 · Pro Max $100 · Power $200	Credits: 1,000 / 2,000 / 5,000 / 10,000	Monthly, no rollover	Credits yes; credits per task no	$0.04 per add-on credit
OpenCode Go^[9]	Go $10 · Go Plus $40	Dollar limit per model per month (e.g. $15–$80 on Go)	5-hour = 20%, weekly = 50% of the monthly limit	Yes, in dollars per model	Free models, or opt-in fallback to pay-as-you-go balance
Zed^[10]	Pro $10	$5 of tokens included	Monthly	Yes, in dollars	API list price +10%

Prices are monthly, USD, individual plans. Full plan data, including free and team tiers: all plans or pricing.json.

Claude Code usage limits

Claude Code is included on Pro ($20/mo, or $17/mo billed annually), Max 5x ($100) and Max 20x ($200). The Free plan doesn't include it. Usage resets on a rolling five-hour session window, and paid plans add weekly limits on top. Claude Code and Claude on web, desktop and mobile draw from the same pool, so a long chat session eats into your coding budget.^[1]^[2]

Anthropic publishes only ratios: Pro is at least 5x Free per session, and Max is 5x or 20x Pro. When you hit the cap you can keep going with usage credits at standard API rates, or switch Claude Code to API pay-as-you-go. Details: Claude Code limits explained.

Cursor usage limits

New Cursor plans (Plus $8, Premium $30, Super $100, Ultra $200) include a weekly pool for Cursor's own models, Grok 4.7/4.6/4.5 and Composer 2.5, shared with the Grok apps. Premium is 4x Plus, Super 15x and Ultra 50x.^[3] Third-party models aren't in the pool. On Premium and above they bill “at the model's API list price from the first request” from prepaid credit, and Plus can't use them at all.^[4] So for Claude-heavy users, the Cursor “limit” is really your top-up budget. Legacy Pro/Pro+/Ultra keep their old monthly pools until you switch. More: Cursor limits and what changed in October.

Codex limits on ChatGPT Plus and Pro

OpenAI is the only big vendor that publishes numbers, though as wide ranges. On Plus, the local messages per 5 hours are: GPT-6 Astra 5–45, GPT-6.1 Sol 15–160, GPT-6 Sol 15–150, GPT-6 Luna 350–3,000. Plus also has weekly limits. Pro ($100, $200 or $500) currently has no 5-hour limit, though weekly limits may apply. Past the limit you buy ChatGPT credits or use Codex with an API key. Fast mode uses limits faster.^[5] The page doesn't say what a credit costs in dollars, so we don't convert it. If you've seen claims that a top Pro tier “ran out in a day”, those are individual reports; OpenAI's page only commits to the ranges above. More: Codex limits.

GitHub Copilot: limits in dollars

Copilot meters agent and chat work in AI credits worth $0.01 each: Pro $10 includes $15 of credits, Pro+ $39 includes $70, Max $100 includes $200. Completions and next-edit suggestions are unlimited on paid plans and don't use credits. Past the allowance you set a dollar budget.^[6] GitHub doesn't publish how many credits each model burns per request, but at least the ceiling is in dollars. More: Copilot credits explained.

Devin Desktop, Kiro, OpenCode Go and Zed

Devin Desktop (formerly Windsurf) replaced credits with daily and weekly quotas in March 2026 and doesn't publish the numbers. Extra usage bills at API pricing.^[7] See also our note on SWE-2's free window.
Kiro sells monthly credits (1,000 on Pro $20 up to 10,000 on Power $200) with add-ons at $0.04. It doesn't say how many credits a task costs.^[8]
OpenCode Go ($10) and Go Plus ($40) set a dollar limit per model per month, for example $15 for Grok 4.7 or Claude Haiku 5.5 on Go and $60 on Go Plus. Each 5-hour window can use 20% of that and each week 50%. It covers open-weight and low-cost models, not Claude Sonnet/Opus.^[9] See OpenCode pricing.
Zed Pro ($10) includes $5 of tokens, then charges API list price +10%.^[10]

How to pick a plan by limits

You code in long bursts: 5-hour windows (Claude, Codex Plus) bite first. Max 5x/20x or Codex Pro move the ceiling up, and the weekly cap becomes the real limit.
You want a predictable bill: pick a dollar-metered plan (Copilot, Zed, OpenCode Go), or set a hard monthly cap on Cursor top-ups.
You live on Claude Sonnet/Opus: compare Claude Pro/Max flat pricing against API rates through Cursor Premium. Our calculator runs both for your hours per day. Head-to-head: Cursor vs Claude Code, Claude Code vs Codex.
You bring your own API key: there's no plan limit, only your provider bill. Set a provider-side spend cap. Our BYOK safety guide shows how for OpenAI, Anthropic and Google.

FAQ

What is the Claude Code usage limit?

Claude Code limits reset on a rolling five-hour session window, with weekly limits on top for paid plans. Anthropic publishes only relative sizes: Pro is at least 5x Free per session, and Max is 5x or 20x Pro. Claude Code and the Claude apps share one pool. There is no published message or token number.

What is the Cursor usage limit on the new plans?

Since October 12, 2026 Cursor plans include a weekly pool for Cursor models (Grok and Composer), shared with Grok. Premium is 4x Plus, Super 15x and Ultra 50x. Claude, GPT and Gemini are not in the pool; on Premium and above they bill at API list price from prepaid credit.

Does ChatGPT Pro have a Codex limit?

OpenAI says Pro currently has no 5-hour limit, but weekly limits may apply. ChatGPT Plus has published ranges of local messages per 5 hours that depend on the model, for example 15–160 with GPT-6.1 Sol.

Which AI coding plan has the most transparent limits?

The plans that meter in dollars: GitHub Copilot (AI credits at $0.01), OpenCode Go (a dollar limit per model) and Zed Pro ($5 of tokens, then API +10%). Claude, Cursor and Devin publish multipliers or nothing, so you only learn the real size by hitting it.

What should I do if I keep hitting my limit?

Check which window you hit (5-hour vs weekly), move routine work to a cheaper model, and compare the next tier against paying API rates for the overflow. Our calculator estimates both for your usage.

Sources

X counts come from the X API recent-counts endpoint for our usage-limit keyword query (English, excluding reposts), full UTC days October 14–19, 2026. Counts include any post using those terms and aren't a measure of sentiment.

Estimate your own monthly cost → · All limits pages · Pricing changelog

Get pricing-change alerts

Email me when Cursor, Claude, Codex or Copilot change prices or limits. No spam, roughly 1–4 emails a month.

We only use your email for these alerts — never sold or shared. Stored on Cloudflare. See our privacy note.