ChatGPT free version limitations next to Claude and Gemini

A free limit is not one thing

The pages we read use the word limit for five different things, and mixing them up is an easy way to describe a free plan wrongly. The first is a cap on text itself. OpenAI’s Free Tier FAQ says Free users have unlimited everyday text chats, subject to abuse-prevention safeguards, and its pricing page repeats this with the phrase “abuse guardrails”. The second is a window of time. Anthropic’s Get started page says the free Claude plan has a session-based usage limit that resets every five hours, and Google’s page on Gemini Apps says its limit refreshes every five hours until you reach a weekly limit.

The third is that weekly limit. Anthropic’s article on usage limits lists a weekly limit under paid plans, and the free-plan text we read mentions the session limit only. Google’s opening sentence names the weekly limit without separating the plans. The fourth is a separate limit on a tool: OpenAI names data analysis, file and image uploads and image creation as tools whose limits can be separate from the main chat limit. The fifth is the length of one conversation. Anthropic calls it a length limit and says usage limits are about quantity over time, while length limits are about the depth and complexity of individual conversations. OpenAI describes the same idea through two notices, “Context window full” and “Chat memory full”.

Each page we read describes one provider at a time. They do not set the three free plans side by side, and they do not say which of the three runs out first, so this page keeps the three apart and attributes each statement to the page that makes it.

ChatGPT: the text is open, the tools are not

OpenAI’s Free Tier FAQ, which showed an update time of 15 hours before we read it on 30 October 2026, says ChatGPT is free to use and that the default model and available limits can change over time. On rate limits it is direct: Free users have unlimited everyday text chats, subject to abuse-prevention safeguards, while file uploads, image generation, voice, data analysis and other tools have separate usage limits, and ChatGPT notifies you when you reach one. If you reach the limit on data analysis, uploads or image creation, the FAQ says you will need to wait for a later time. It says GPT use follows your current Free model limit, that access pauses until the limit resets, and that ChatGPT shows the reset timing in the product. The FAQ does not give a number of messages, hours or days for any of these.

The same FAQ lists what a Free user can do: search the web, analyze data, upload images or files, discover and use GPTs, create images and use a Library with 500 MB of storage. Creating or publishing a new GPT is not available on personal accounts, which it says includes Free, Go, Plus and Pro. If you hit a Free limit, upgrading to Plus, Pro or Business resets your usage rates, according to the FAQ.

Help Net Security reported on 7 August 2026 that OpenAI was removing the rate limit on text chats for free users “starting next week”, that unlimited chats would stay subject to what OpenAI calls “abuse guardrails”, and that limits would remain for file uploads, images and other tools. The article adds that Free and Go users were being moved to GPT-5.6 Luna and getting a Think button. The FAQ we read on 30 October matches it on text and tools. The pricing page’s Free column reads “Unlimited text chats with GPT-6” and then lists limited messages with uploads, limited image creation, limited voice chats, limited deep research, limited memory and context, limited Codex access and limited ChatGPT Work access on desktop.

The OpenAI pages do not agree on the model’s name. The FAQ and Help Net Security say GPT-5.6 Luna. The pricing page says GPT-6 Luna is included in Free, Go and Plus, yet its model table also has a GPT-5.6 Luna row in which Free is marked No, and Android Police writes that ChatGPT offers GPT-6 Luna. We cannot say which is current. The FAQ says ads may appear in the Free tier in certain countries, and the pricing page says the Go plan may include ads; our page on whether ChatGPT is free covers ads and what OpenAI says about your chats.

Claude and Gemini: a five-hour window and no message count

For free Claude, Anthropic’s Get started page says there is a session-based usage limit that resets every five hours, that the number of messages you can send “will vary based on demand”, and that Anthropic may impose other types of usage limits. Claude notifies you when you reach the limit or when a prompt exceeds the available context window. The page gives no number of messages. Anthropic’s article on usage and length limits says that on reaching a session or weekly limit you can wait for the reset or upgrade, that a new plan’s limits apply right away, and that with usage credits enabled your work continues on credits. In the pricing table, usage credits are marked No for Free and Yes for Pro. The same table marks Opus No on Free and Yes on Pro, with Sonnet and Haiku Yes in every column; Android Police, in a passing comparison, writes that Claude gives free users only its mid-tier Sonnet, which does not match the table’s Haiku row.

For Gemini, Google’s page says the limits are compute-based and depend on the complexity of your prompt, the models and features you use, and the length of your chat. A table sorts the cost of a request: a simple chat is very low, an ongoing conversation and large file analysis are medium, and Deep Research and video generation are high. The page says Gemini Apps notifies you when you are close to a limit and, at the limit, tells you when it will refresh; the usage limits are in Settings. For people without a Google AI plan it warns that features needing more compute, such as Deep Research, may be unavailable in periods of high demand, that their limits may be cut before those of subscribers, and that limits may change without notice.

Android Police, in an article published on 27 October 2026, writes that since 9 October free accounts are limited to Gemini 3.5 Flash-Lite. It says you can still pick a low, medium or high thinking level on Free but that the allowance runs out faster at the higher ones, and that Google does not provide exact usage limits, so there is no set number of prompts. Heavier features, including Deep Research and image generation, use the allowance faster, in its words.

How much each can read at once

The OpenAI pricing page, Google’s Gemini Apps page and Anthropic’s pricing page each print a size for how much one conversation can hold, and they do not use the same unit. OpenAI’s pricing page gives GPT Instant a total context window of 27K on Free and 54K on Go and Plus, with an input maximum of about 12 pages of text on Free and about 40 on Go and Plus. For its reasoning models the Free column says “Varies”, against 256K for Go and Plus. The page does not say whether K counts tokens, so we do not either. Google’s page lists 32k tokens without an AI plan, 128k with AI Plus and 1 million with AI Pro and AI Ultra, and likens the context window to reading capacity. Claude’s pricing table prints “Up to 1M varies by model” in the Free, Pro and both Max columns, while Anthropic’s support article says that on paid plans the newest models support up to a 1M token window and others 500K or 200K. We read the two Anthropic pages as differing: the article speaks of paid plans and the table prints the same text in every column.

Android Police’s author tested the free Gemini chat box with a handbook of about 124 pages. It accepted about 5,000 words, roughly 10 pages, before cutting off mid-sentence, and the article gives uploading the text as a file as the workaround. That is the author’s test, not ours.

A full chat is a related limit. OpenAI says your messages and ChatGPT’s replies both take up space, that a few long exchanges can take more room than many short ones, and that the number of messages per chat is therefore not fixed. On Free you can start a new chat right away, and if the notice offers an upgrade you can continue in the same chat. A new chat started with a summary may leave out an earlier instruction or the exact wording of something you wrote. Anthropic says that with code execution on, Claude summarizes earlier messages as a conversation nears the window, and that longer conversations that trigger this use more of your usage limit. Google’s tip is to start a new chat when you move to a new topic, so Gemini does not have to re-read a long history.

What the money buys

OpenAI’s page on Plus gives $20 a month, billed monthly, and lists priority access during high-traffic periods, higher model limits, which it describes as more messages and broader model options than Free, and faster responses. The same page says Plus subscriptions may include usage limits such as message caps, especially during high demand. The pricing page, which showed euros when we read it from Europe, lists Free at 0, Go at 8, Plus at 23 and Pro from 103 a month. It marks Plus “Limits apply” where the Free and Go columns carry the abuse guardrails line, and adds projects, scheduled tasks and custom GPTs, early access to new features and expanded memory and context. The FAQ says creating a new GPT is unavailable on personal accounts including Plus, so the pricing page and the FAQ read differently on custom GPTs. OpenAI’s pages that we read do not give a multiplier for Plus.

Anthropic’s support article puts Pro at $20 a month in the United States and lists more usage per session than Free, priority access during high-traffic periods and early access to new features. The pricing page shows $17 a month with an annual subscription ($200 billed up front) or $20 billed monthly, and says Pro adds Claude Code, Claude Science, Projects, more Claude models and Claude in Chrome and Microsoft 365. Free is marked as having Projects “Up to 5”. The Pro card reads “More usage” with no multiplier, but the FAQ lower on the same page says Pro gives at least 5x more usage per 5-hour session than Free; the 5x and 20x figures for Max are measured against Pro, and the page puts Max at $100 and up.

Google’s table puts AI Plus at 2x the standard limits, AI Pro at 4x and AI Ultra at 5x or 20x of AI Pro. Android Police says AI Plus costs $5 a month and adds Gemini in Gmail, while Docs and Sheets need Pro, and that the free tier does not work inside Gmail, Docs and Sheets. It adds that Pro brings 5TB of storage against 15GB free, YouTube Premium Lite, Google Home Premium Standard and 1,000 monthly credits in Flow, and that for someone who pays for those services Pro would cost under $20 in practice. Its headline puts Pro at $20. The Google page we read does not give prices. Our pages on whether Claude is free and whether Gemini is free go through each plan, and Go or Plus compares OpenAI’s two paid steps.

Who may need to pay, and what the pages leave out

Only one of the pages we read gives a view on who should pay, and it is about Gemini. Android Police writes that the free tier is more than enough for most everyday tasks, that Pro makes sense if you want Gemini inside Docs and Sheets, want to make videos, hit usage limits often, or pay for related services, and that the $5 AI Plus plan is a cheaper middle ground for someone who wants Gemini in Gmail and a better model than Flash-Lite. The OpenAI and Anthropic pages we read mostly describe what each plan includes; Anthropic’s pricing page says only that Free covers everyday questions and that Max is built for people who work with Claude all day. For what to do at a limit, Anthropic and Google both say to wait for the reset or upgrade, and OpenAI says an upgrade resets a reached Free limit.

The pages we read do not say:

• how many messages, hours or days apply to ChatGPT’s tool limits, or what the upload cap is;
• how many messages free Claude allows, or when the five-hour window starts (Anthropic’s pricing page says paid plans add weekly limits on top of the five-hour window and does not describe a weekly limit for Free);
• how many prompts free Gemini allows, how long its weekly cap lasts, or when it resets;
• the dollar price of ChatGPT Go and Pro, which we saw only in euros, or a Gemini AI Pro price in Google’s own page.

The tick marks in Google’s table of feature availability did not come through in the text we read, so we do not say from it which feature sits in which plan. We opened no account and measured no limit; every statement here is a statement of the page named.