Free AI Chat: Paid Plans and Limits Compared

All three vendors list no-cost entry options, but their paid plans use different limits and feature allowances.

Sources checked 2 Oct 2026

Free AI Chat is listed by all three vendors: OpenAI’s ChatGPT pricing page, Anthropic’s Claude pricing page, and Google’s U.S. AI plans page each show a $0 starting option, as read on October 2026. Paid plans add different allowances and features, but all retain some form of usage limit. Chat Picker has not tested the assistants for your tasks; this page compares only what their pages document.

What the vendors document

For Free AI Chat, compare the limit rule and included tools, not just the monthly price.

ChatGPT: OpenAI lists unlimited text chats on Free, with limited uploads, images, voice, and Deep Research. Its Free Tier FAQ also lists web search, file and image prompts, image creation, and 500 MB of Library storage. OpenAI says the default model and available limits can change, while ads may appear in certain countries.
Claude: Anthropic lists Free at $0 for chat on the web, desktop, and mobile, plus web search, file creation, and code execution. Its pricing page says limits reset on a rolling five-hour window, paid plans add weekly limits, and there is no fixed message count. Capacity varies with conversation length and complexity, model choice, and feature use.
Gemini: Google lists Free at $0 with a Google Account, including image generation and editing, Gemini Live, and 15 GB of storage. Google’s Gemini Apps limits page says capacity depends on prompt complexity, models, features, and chat length; it refreshes every five hours until the weekly limit is reached.

For an individual paid step, OpenAI lists ChatGPT Plus at $20 per month, with more messages, broader model options, voice, image generation, file analysis, and Deep Research where available. Usage caps can vary with system conditions. OpenAI’s pricing page showed Go at A$13 per month on an Australian view and says it may include ads; its Go Help page provides no U.S.-dollar price.

Anthropic’s captured pricing page listed Claude Pro at $34 monthly or $28 per month with annual billing, with $340 due up front. Its Pro Help article instead listed $20 monthly in the U.S. Because the vendor pages do not agree on one universal figure, confirm the regional checkout price. Google’s U.S. page listed AI Plus at $4.99 per month with 2x Free usage and 400 GB of storage, and AI Pro at $19.99 per month with 4x Free usage and 5 TB.

Training choices are documented differently. OpenAI says an opt-out from training on chats is available for Free, Go, Plus, and Pro; Anthropic says it is available for Free, Pro, and Max, while Team is not trained on by default. Google’s plans page does not state a training setting. These are training controls, not retention promises: the cited plan pages do not settle complete retention or deletion schedules.

The accuracy notes set direct limits on trust. OpenAI’s accuracy note says ChatGPT can be incorrect or misleading and may sound confident when wrong. Anthropic’s note says Claude can also produce misleading statements and should not be the only source of truth. Google’s safety guidelines say Gemini should not generate factually inaccurate outputs that could cause significant harm to health, safety, or finances, while also considering context. None of these statements guarantees that an answer will be correct.

What the documentation cannot tell you

The pages do not tell you how a particular prompt will perform, how quickly an account will respond under your conditions, or which model you will receive if availability changes. A plan multiple also does not reveal a fixed number of completed tasks.

Your own trial can show how an assistant handles the documents, tools, and questions you actually use. It does not establish a permanent advantage over another assistant.

How to check it yourself

Context: Give the assistant a PDF report you have permission to use. Ask: “Create a chronology from this report and identify every date, version number, and unresolved issue.” Record omissions, conflicting details, and whether the full file was accepted.
Model label: Ask: “State the model label shown in your interface, quote it exactly, and do not guess if no label is visible.” Repeat after changing the visible model option. Record the labels shown and any availability change.
Files and audio: Create an image containing the sentence “Project Cedar is a fictional neighborhood energy project.” Upload it, then submit the same request by text, image, and voice: “Explain what benefits and limitations a community energy project might have.” Record transcription problems, missed image details, and plan restrictions.
Limits and speed: Use the prompt “Explain why libraries use call numbers.” Keep network conditions and conversation length consistent. Record the start time, elapsed time, task complexity, any cap message, and the reset information shown. Treat one run as an observation, not a speed result.
Data controls: Upload only the fictional Project Cedar text. Before doing so, inspect the account’s training, retention, and deletion controls. Record the exact labels you find. If the interface does not explain them, mark that point unresolved rather than assuming a policy.
Price and value: Use the report from step one and ask: “Summarize this report for a new employee and list unresolved questions.” Before subscribing, record the displayed price, billing cadence, annual charge, and stated limits. Note which features you used and whether the answer met your standard.

Which rows of the comparison matter

In the Claude vs ChatGPT vs Gemini comparison matrix, prioritize these rows:

Free plan
Low-cost tier
Main paid plan
Heavy-use plans
Context window in the app
How usage limits are described
Training on your chats
Ads
Storage and multimodal access

Treat “Not verified” as missing information, not as zero access or permission. Check each vendor link and its read date before subscribing, especially where regional pricing or changing model availability appears.