ChatGPT Plus vs Claude Pro for coding: the $20 plans

We make Shuriken. We tried to be fair; corrections welcome at office@shuriken.sh. We don't sell model usage, and neither company pays us.

Quick answer#

ChatGPT Plus and Claude Pro both cost $20 a month, and for coding each one is really a ticket to its company's agent. Claude Pro gets you Claude Code, metered by a five-hour session limit and a weekly limit that you share with the Claude chat app. ChatGPT Plus gets you Codex, metered per five hours, with OpenAI's message estimates changing a lot by model, plus weekly limits that "may also apply". Neither company publishes a hard number, so the better plan is the one whose agent does better on your code.

At a glance#

We read both vendors' pages on 24 October 2026. Prices are US dollars before tax and can differ by country and in mobile app stores.

How the limits work on each plan#

The prices match. The meters don't, and the meter is what you'll notice on a Tuesday afternoon.

Claude Pro: one allowance, two clocks#

Anthropic's Claude Code help page says it in one line: "Both Pro and Max plans have a five-hour session limit and a weekly limit." Both are shared with Claude on the web, desktop and phone, so a long research chat in the morning leaves less for the terminal in the afternoon. Every message counts toward both clocks at once.

Anthropic doesn't say how many prompts fit in a Pro session. Its usage help lists what moves the number instead: conversation length and complexity, the features you use, the model and the effort level. The one ratio on the pricing page is against Free: Pro gives "at least 5x more usage per 5-hour session". Our 5-hour limit guide explains when the window starts and resets.

If the limit lands while Claude Code is partway through a response, the Claude Code help page says it "may keep working briefly so it can finish the step in progress" and shows "Usage limit reached · wrapping up". On Pro that allowance is small, applies up to once per weekly usage period, and is still rolling out. After that, Claude Code says so plainly ("You've hit your session limit · resets 3:45pm") and, per its error reference, "blocks further requests until the reset time shown in the message". In an open interactive session it can wait and carry on with the interrupted task shortly after the reset. Or you turn on usage credits with /usage-credits and pay standard API rates for the overflow.

ChatGPT Plus: five-hour estimates by model#

OpenAI's Codex pricing page is the only one of the two that prints numbers. These are its estimates of local messages per five-hour period on Plus, read on 24 October 2026:

OpenAI explains the spread: small scripts may use a fraction of the allowance, while larger projects and long-running tasks "use significantly more per message". It adds that "these estimates are not fixed message limits" and that cloud tasks "may use more of your allowance than local messages". Weekly limits may apply on top, with no figure given. The page also says ChatGPT Work and Codex share usage.

At the wall, Codex is more lenient with the turn in progress. OpenAI says that if you reach the limit during an active turn, "the agent will be able to continue working on that turn, subject to fair use limits." After that, Plus users can buy credits (the price depends on your plan, and the page doesn't list it) or run extra local chats on an API key at API rates. /status in the CLI shows what's left. Our Codex usage limits guide has the 2026 history of these limits, which moved more than once.

Cloud work counts too#

Both plans let the agent run in the vendor's cloud, and both bill that to the same allowance. Anthropic's cloud sessions page says cloud sessions are available on Pro and count "toward your plan's usage limits alongside the rest of your Claude and Claude Code usage". OpenAI says local messages and cloud chats share the Plus allowance.

Which models each $20 plan gets you#

On Pro, Anthropic's pricing table lists Opus, Sonnet and Haiku. Fable is the exception: Anthropic's Fable help page says Fable 5 and 5.1 "aren't included in your plan's usage limits" on Pro and run on usage credits from the start. On Max, Fable is part of the plan, up to half the weekly limit.

On Plus, OpenAI's plan card names GPT-6.1 Sol and GPT-6 Luna, its feature table ticks GPT-6 Sol as well, and the usage table above also gives Plus a range for GPT-6 Astra. Fast mode is ticked for Plus; Ultrafast is not (on a personal plan it needs Pro at $500).

Model names in this market change every few months. Treat this section as a reading from 24 October 2026, not as a rule.

ChatGPT Plus vs Claude Pro for coding: which one?#

We won't name a winner, because it would be wrong after the next model release. What decides it for most people:

1. Which agent does better on your code. Give Claude Code and Codex the same three tasks from your real backlog. Our Claude Code vs Codex comparison covers the agents themselves: install, MCP config, licence and platforms.
2. Which meter fits your day. Claude Pro is one allowance per five hours plus a week, shared with your Claude chats. Plus has a five-hour allowance that OpenAI estimates in messages, and how many fit depends heavily on the model.
3. What happens at the wall. Claude Code may finish the step it's on, then stops new requests until the reset. Codex can keep working on the turn it's on, subject to fair use. Both let you pay for more: usage credits on Claude, ChatGPT credits or an API key on Codex.
4. What else you use. Pro also covers Claude on the web and phone; Plus covers the rest of ChatGPT. If you already pay for one for chat, that agent costs you nothing extra to try.

When $20 stops being enough#

OpenAI's plan card pitches Plus as a way to "power a few focused coding sessions each week"; Anthropic gives Pro no such line. If you run out most days, or run two or three agents at once, the next step on both sides starts at $100 a month: Claude Max (5x or 20x Pro's per-session usage) or ChatGPT Pro (which, OpenAI says, currently has no five-hour limit). We compare those in Claude Max vs ChatGPT Pro for coding, and Claude Pro vs Max for Claude Code has the signs you've outgrown Pro. For Copilot, Cursor and Google's plans next to these two, see the best AI coding subscription page.

Holding both#

If you code with agents every day, one of each costs $40 a month. You get two agents and two sets of limits from two companies, and one running out doesn't touch the other.

The cost of that is two worlds: two terminals, two MCP configs, two places to check what's left. In Shuriken, each ChatGPT or Claude sign-in you add is a source, and one project can run a Claude Code chat next to a Codex chat. One list of MCP servers serves both agents, and the Sources page shows each account's 5-hour and weekly use side by side.

If you hold more than one account with the same provider, hot swapping helps as well: Shuriken moves your chats to another of your accounts or sources when one nears its limit (99% by default). A Claude Code chat moves on its next request; a Codex chat moves when its current turn ends. A chat only ever moves to another source of its own agent, so a Claude Code chat never turns into a Codex chat, and moving work from one agent to the other stays your call. The source switching page has the details. Only add accounts you own and are entitled to use; providers' terms may not allow sharing a subscription sign-in.

Shuriken never bills model usage, and its Free plan holds two sources, enough for one Claude sign-in and one ChatGPT sign-in. If you pay for both, download Shuriken and run them in one place.