Nano GPT logo
NanoGPT

NanoGPT Accessibility Statement

Last Updated: October 21, 2026

NanoGPT LLC wants NanoGPT to work for as many people as possible, including people who use a screen reader, screen magnifier, speech recognition, switch access, or a keyboard instead of a mouse. This statement covers the NanoGPT website and web app at nano-gpt.com: chat, image, video and audio generation, sign-in, payments and subscriptions, and our API documentation.

About NanoGPT

NanoGPT gives you access to many AI models in one place. You type or dictate a message, choose a model, and the answer or the image, video, or audio you asked for appears in the conversation. You can start without an account and sign in later. You pay as you go from a prepaid balance or with a monthly subscription, and developers can use the same models through our API.

Our Goal and Current Status

We are working toward the Web Content Accessibility Guidelines (WCAG) 2.2 at Level AA. NanoGPT partially meets this standard: some parts do not yet fully meet it. The main gaps we know of are listed below, with a way around each one where there is one.

What We Have Done So Far

  • Keyboard access and focus handling in dialogs, menus, and the phone navigation, with focus returned to where you were when they close.
  • Labels for buttons and form fields, including the message box, settings, and voice controls.
  • A heading structure on our main pages, and improved color contrast, including in the light theme.
  • Keyboard reordering of queued messages: press Space on a message's handle to pick it up, use the arrow keys to move it, and press Space again to drop it.

Features That May Help

  • Light, dark, and system color themes, from the quick settings in the sidebar.
  • Five text sizes, under Font size in Settings.
  • A setting for whether Enter sends your message or starts a new line.
  • Voice input for dictating messages, and a voice mode that also shows a text transcript.
  • Several ways to sign in (an email code, a passkey, or Google or GitHub) and to pay (card, bank payment, or cryptocurrency), so you can choose one that works with your setup.
  • In the current chat view, screen readers announce "Response finished" when an answer is complete, instead of reading every word as it streams in. The older conversation view doesn't do this yet.
  • Image descriptions: ask a model that can read images to describe an attached or generated image.
  • Our API, which lets you use NanoGPT models from any compatible app or your own scripts, if the website doesn't suit your setup.

Known Limitations

  • AI-generated images, diagrams, and video: these come without a written description, and generated video has no captions or audio description in our player. You can ask a model that can read images to describe an image, or ask for a text explanation instead of a diagram.
  • AI-written answers: the model decides how an answer is structured (headings, lists, tables), so structure can be inconsistent. You can ask the model for a simpler format.
  • Charts: charts on the Usage, Leaderboards, and Benchmarks pages show exact values only when you hover with a mouse, and can't be explored with a keyboard. The lists and tables on those pages carry much of the same information.
  • Drawing on images: painting a mask for image editing and drawing highlights on an attached image need a mouse, pen, or touch screen. Many image models can edit from a written description instead, but that doesn't give the same precise control as a mask.
  • Reordering queued messages with a mouse or touch screen: this needs dragging, and there are no move up or down buttons yet. The keyboard method above works without dragging.
  • Voice mode transcripts: with some voice models, the text of a spoken reply appears only once the reply has finished, not as it is spoken.
  • Sign-in checks and time limits: requesting a sign-in code by email includes an automated security check, which occasionally asks you to complete a challenge within 2 minutes. Sign-in codes expire after 10 minutes, and you can request a new one. If either gets in your way and your account has another sign-in method, such as Google, GitHub, or a registered passkey, you can use that instead.
  • Other companies' pages: some steps happen on pages run by other companies, such as payment checkouts and the sign-in security check. We don't control their accessibility.
  • Smaller issues found by automated testing: on some buttons, the name read out by assistive technology doesn't match the visible text (for example the model picker, which matters for speech control); a few scrollable code samples and tables can't be scrolled with the keyboard alone; and some controls are smaller than recommended.

Report a Problem or Ask for Help

If something on NanoGPT is hard or impossible for you to use, email support@nano-gpt.com. Putting "Accessibility" in the subject line helps us find it quickly. Please include:

  • the page or feature, and its web address if you have it;
  • what you were trying to do, and what happened;
  • if you like, your device, browser, and any assistive technology you use.

You don't need to tell us about any disability. Never send passwords, API keys, or card numbers.

We can't promise to fix a particular problem or to fix it by a particular date. You can also open a ticket from the Support page.

How We Assessed NanoGPT

This statement is based on our own assessment in October 2026: a review of our source code, and automated testing of key pages on desktop and phone screen sizes with the axe-core accessibility checker. NanoGPT has not yet had an independent accessibility audit.

Technical Details

NanoGPT is built with HTML, CSS, WAI-ARIA, and JavaScript, and needs JavaScript turned on. We design it for current versions of Chrome, Edge, Firefox, and Safari on desktop and mobile.

Changes to This Statement

We may update this statement from time to time. The date at the top shows the latest update. For our other policies, see the Terms of Service and Privacy Policy.