Frequently asked questions

Here are straight answers to the questions people ask most about Llama.Online: how the free chat works, which Llama models it uses, where your conversations are kept, what the tools and assistants do, and how we handle privacy and ads. If your question is not here, send it to us through the contact form.

Using the chat

Do I need an account to use Llama.Online?

No. There is no sign-up, login or subscription. Open the chat, type a message and press Enter. Because there is no account, your conversations are saved in the browser you are using rather than on our server, which also means they do not follow you automatically to another device.

How many messages can I send for free?

Each visitor gets a daily allowance of free messages, currently 100 per day. The allowance is shared between the chat and the tool pages, and the counter in the chat shows how many messages you have left. It resets every day. When you reach the limit, the chat tells you so and you can continue the next day. We use the allowance to keep the service free for everyone and to stop automated abuse.

Which devices and browsers work?

Llama.Online runs in any current browser on Android phones, iPhones, tablets, Windows, macOS and Linux computers. Nothing needs to be installed. On small screens the chat history moves into a menu so the conversation gets the full width. Very old browsers may not support some features, such as saving history or attaching images.

Can I stop, edit or regenerate an answer?

Yes. Press Stop while an answer is being written to end it early. You can edit a message you sent and send it again, which creates a new answer from that point. Regenerate asks the model for another version of its last reply, and you can switch between versions. Trying a different model on the same question is another quick way to compare answers.

Why did I see a "model busy" message?

The models run on a shared cloud service, and occasionally one of them is overloaded for a moment. The chat retries once by itself. If the message stays, wait a few seconds and try again, or switch to another model in the menu: Llama 3.3 70B is a good fallback for text questions.

Can the chat search the web, create images or listen to my voice?

No. The Llama models in our chat cannot browse the internet, so they do not see today's news, prices or weather. They write text only and cannot generate pictures. There is no voice input or output. What they can do is read images you attach when you use Llama 4 Maverick or Llama 4 Scout.

Llama models

Which Llama models are in the chat?

Four official models from Meta: Llama 4 Maverick, the default; Llama 4 Scout; Llama 3.3 70B; and Llama 3.1 8B. They are open-weight models, which means Meta publishes the trained model files. We do not host them ourselves; a cloud inference provider runs them, and every answer is labelled with the model that wrote it.

Which model should I choose?

Keep Llama 4 Maverick for most things: writing, explanations, coding help and questions about images. Choose Llama 4 Scout when you paste a very long document, since it accepts the longest input. Llama 3.3 70B is a reliable text-only alternative, and Llama 3.1 8B is the fastest for short, simple tasks. Our guide to choosing the right Llama model goes into more detail.

How do I send an image to Llama?

Select Llama 4 Maverick or Llama 4 Scout, then attach an image with the image button, paste it from the clipboard or drag it into the chat. You can add one image per message, in JPEG, PNG or WebP format; large photos are made smaller in your browser before sending. If a text-only model is selected when you add an image, the chat switches to Maverick automatically.

Are these the real Meta Llama models?

Yes. The chat sends your messages to the official Llama 4 Maverick, Llama 4 Scout, Llama 3.3 70B Instruct and Llama 3.1 8B Instruct models, hosted by a cloud provider. We add a short system instruction, for example to answer in your language, but we do not fine-tune or alter the models. Answers can still differ slightly from other services because of settings such as answer length.

What is the newest Llama model, and why are Meta's newer models missing?

The newest models with the Llama name are Llama 4 Maverick and Llama 4 Scout, released in April 2025. A larger Llama 4 Behemoth was announced at the time but has not been released. Since 2026, Meta has published its new models under a different name, Muse; some are open-weight and others are available only through Meta's own services. They are not part of our chat. The Llama history timeline has the full story.

How up to date is Llama's knowledge?

According to Meta's model cards, the training data of Llama 4 Maverick and Scout ends in August 2024, and that of Llama 3.3 70B and Llama 3.1 8B in December 2023. The models know nothing about later events unless you paste the information into the chat. For anything recent or important, check a reliable source. Our guide on fact-checking AI answers shows how.

History and sharing

Where are my conversations saved?

Only in your browser, on the device you are using. We do not keep a copy of your chat history on our server. You can search, rename, pin and delete chats in the sidebar. Clearing your browser's site data, or closing a private browsing window, removes the history, so export any conversation you want to keep.

How do I move my chats to another device?

Open the sidebar menu and choose Export all chats to download a JSON file. On the other device, open the chat and choose Import chats, then pick that file. You can also export a single conversation as Markdown, plain text or JSON, which is handy for pasting it into a document or keeping notes.

How do share links work?

Press Share in a conversation to publish a read-only copy at an address like llama.online/s/ followed by a short code. Anyone with the link can read it. The page is not indexed by search engines, and the copy cannot be edited afterwards, so later messages in your chat are not added. Images are not included in shared copies. Only share conversations without personal details.

How do I remove a conversation I shared?

Send us the share link through the contact form and ask for it to be removed. Deleting the chat in your own browser does not delete the public copy, because the shared version is stored separately on our server. If you find a shared conversation that breaks the rules, use the Report link on that page.

Tools and assistants

What is the difference between the chat and the tool pages?

The chat is open-ended: you talk with Llama about anything and keep the conversation. Each of the twelve AI tools does one job, such as summarizing, paraphrasing or translating, with a form and options tuned for that task. Tool results can be copied or moved into the chat to continue. Both use Llama 4 Maverick and count towards the same daily allowance.

What are assistants?

Assistants are saved instructions that give Llama a role and a way of working, such as Study Tutor, Editor, Travel Planner or Mock Interviewer. When you start a chat with one, Llama follows its instructions until you switch back. There are sixteen built-in assistants, and each comes with a few example questions to get you started.

Can I create my own assistant?

Yes. In the chat, open Assistants and choose Create assistant. Give it a name, a short description, instructions that describe how Llama should behave, and optional conversation starters. Your assistants are stored in your browser along with your chats, and they are included when you export your data.

Where can I find good prompts?

The prompt library has ready-made templates for writing, study, coding, business, marketing, careers, creative work and everyday life. Fill in the parts in square brackets and send. The same library is available inside the chat. To write your own, read our guide on how to write prompts.

Privacy, ads and safety

Do you store or read my messages?

We do not store the text of your messages. To create an answer, each message travels through our server and AI gateway to the inference provider that runs the model. The gateway keeps technical logs, such as the model used, the request size, the time and a hashed form of your IP address. Shared conversations and contact form messages are the exceptions you choose yourself. Details are in the privacy policy.

Why are there ads on the site?

Every answer costs us money to generate, and ads are how we keep Llama.Online free without accounts or subscriptions. Ads are shown around the page content by our advertising partner, never inside the model's replies. In some regions you will see a consent banner first, where you can choose how your data is used for advertising.

Can I rely on Llama for health, legal or money questions?

Use it for general information and to prepare questions, not for decisions. Language models can sound confident and still be wrong, and they do not know your full situation. For medical, legal or financial matters, talk to a qualified professional, and in an emergency contact your local emergency services.

Is Llama.Online affiliated with Meta?

No. Llama.Online is an independent website, and it is not run, endorsed or sponsored by Meta. Llama is a trademark of Meta Platforms, Inc., and we use the name to describe the models in our chat. As Meta's license requires for products that use Llama, we show "Built with Llama". Read more about us or about the Llama license.