2026 AI Tools: Features People Ask About and Documented Limits
This page turns the vendor-documented features and limits of ChatGPT, Claude, and Gemini into a self-run test plan you can complete in an afternoon.
Sources checked 3 Oct 2026
The features people ask about—code generation, writing, data analysis, multimodal input, pricing, and support—are documented on each vendor's own pricing and help pages, not measured by us. Chat Picker has not run quality, speed, accuracy, or benchmark tests of any assistant; the steps below use only what the vendors publish. Older versions of this article contained statistics and test results without sources, and they have been removed.
Know these limits before you start
Plan prices and limits shape what you can test. ChatGPT Free includes unlimited text chats with GPT-5.6 Luna and limited uploads, images, voice, and deep research, while ChatGPT Plus is $20 per month as read on October 2026 (OpenAI's ChatGPT pricing page). Claude Free is $0 and includes chat on web, desktop, and mobile plus web search, file creation, and code execution; Claude Pro is $20 per month billed monthly or $17 on annual billing (Anthropic's Claude pricing page). Gemini Free is $0 with a Google Account and 15 GB storage; Google AI Pro is $19.99 per month and Google AI Plus is $4.99 per month (Google's Gemini plans page).
Context windows differ. ChatGPT lists 27K tokens on Free, 54K on Go and Plus, and 128K on Pro for instant models, with higher figures for reasoning models (OpenAI's ChatGPT pricing page). Claude lists up to 1M tokens on every plan (Anthropic's Claude pricing page). Gemini's context window is 32k without an AI plan, 128k on AI Plus, and 1 million on AI Pro and AI Ultra (Google's Gemini Apps limits page).
Usage limits are described, not counted. OpenAI publishes no exact message counts on its pricing page; Claude states Pro has "more usage" than Free and Max is 5x or 20x Pro, with no message counts; Gemini uses compute-based limits that refresh every five hours up to a weekly cap (Google's Gemini plans page).
Data settings matter for a test. ChatGPT lets Free, Go, Plus, and Pro users opt out of training on chats (OpenAI's data controls help). Claude offers opt-out on Free, Pro, and Max, and says Team is not trained on by default (Anthropic's Claude pricing page). The Gemini plans page we read links to a data-handling notice but states no training opt-out setting, so we do not state one.
Cancellation and export are documented. ChatGPT subscriptions bought on chatgpt.com cancel through the account; uninstalling the app does not cancel (OpenAI's cancellation help). Claude individual users on Free, Pro, and Max can export conversation data through Privacy settings; exports cannot run from iOS or Android apps and the download link expires after 24 hours (Claude's data export help).
Accuracy notes are explicit. OpenAI states ChatGPT can produce incorrect or misleading outputs and calls factual inaccuracies hallucinations (OpenAI's truth help). Claude's help center says misleading responses are a byproduct of current generative AI limits and advises not relying on it as the only source of truth (Claude's incorrect responses help). Gemini shows a Sources button when links are available (Gemini's sources help).
Support channels differ. Claude support is writing-only through a messenger and email, with no phone support (Claude's support help). OpenAI Support starts with a virtual assistant and may escalate to a human (OpenAI's contact support help). Gemini's community page features posts about app features (Gemini community).
The test plan
Run these checks in your own accounts. For each, prepare the input, give the instruction, watch the behavior, and write it in a plain log.
- Code Generation — Prepare a small function request, e.g., "Write a Python function that reads a CSV and returns the average of a column." Give it in the free tier of each assistant. Look at whether code executes or is only shown, and note any plan message about limits. Record the plan, the response type, and any block.
- Content Writing — Prepare a 200-word product description brief with a stated audience. Instruct: "Write a 150-word description for beginner photographers." Look at tone, structure, and whether the tool asks for clarification. Record the output length and any usage cap notice.
- Data Analysis & Visualization — Prepare a sample spreadsheet with 50 rows. Instruct: "Summarize trends and suggest a chart." Look at whether files upload and analyze, and what the assistant returns. Record file size accepted and the kind of output.
- Multimodal Capabilities — Prepare one image and one short text question about it. Instruct: "Describe this image and list three uses." Look at whether the assistant accepts the upload on the plan you use. Record acceptance and any limit message.
- Pricing & Subscription Models — Open each vendor's pricing page and note the free, low-cost, and main paid tiers you see. Look at what features are locked per tier. Record the prices shown and the date you read them.
- Customer Support & Documentation — Open the help center for each and search for "export data." Look at whether the answer is a written article or a contact form. Record the channel and any stated response method.
How to read your results
Your log now holds what each assistant did on the plans you tried. Compare rows side by side: which tasks completed on free tiers, which needed a paid plan, and what each vendor's page says about the limit you hit. No external score is needed. If a task failed because of a documented cap, that is a planning fact, not a quality verdict. Use the log to match a plan to the work you actually do, and re-check the vendor page before subscribing because prices and limits change.
Where the plans differ
The comparison matrix at Claude vs ChatGPT vs Gemini lists the rows that separate the assistants: free plan availability, low-cost tier presence, main paid price, heavy-use plans, team plan, context window in the app, how usage limits are described, training on your chats, and ads. Those rows are drawn from the same vendor pages cited above. Chat Picker marks any figure it could not confirm as "Not verified" and leaves it blank, so check the matrix cell and its source date when a decision depends on it.
Sources: - ChatGPT pricing page - Claude pricing page - Google AI plans - What is ChatGPT Plus? - Claude support help - Gemini Apps limits