WebSLM
Open small language models, right in your browser. There's nothing to install, it works offline, and nothing you type leaves your device. The code is open on GitHub.
11 light models · 150 MB to 1 GB · WebGPU · free and open source
A real session: Liquid LFM2.5 1.2B on a laptop with integrated graphics, 8.5 tokens per second, all on the device.
WebSLM is a free, open way to run small language models on your own computer. Private by default, fast enough for everyday writing, and yours to inspect.
A full chat app in a browser tab, built for everyday writing and reading, with no account needed to keep your words.
Attach a PDF, a Word document or a text file and ask questions about it. Your browser reads the file. It's never uploaded anywhere.
Models from Google, Microsoft, Meta, IBM, Mistral AI, Alibaba Qwen, Liquid AI and Hugging Face, from 150 MB to 1 GB, so they run on most devices. Every file is verified before it's offered.
Pick up where you left off. Chats are saved on this device only, with search, rename and delete, and you can export them to a file whenever you like.
Why WebSLM
Using an AI assistant today usually means sending your text to a company's servers. For a client file, a medical note, a tenant's letter or an employee review, that's a step many people shouldn't take, and many workplaces don't allow it.
Models from a few hundred million to a few billion parameters are now genuinely useful for rewriting, summarizing, translating and drafting, and small enough to run on an ordinary laptop.
WebAssembly and WebGPU let a web page run demanding code at close to native speed. llama.cpp, the open engine behind much of local AI, now runs in the browser.
Open a page, pick a model and chat. The model runs on your machine, with a free sign-in before your first download. Because the code is open, you can check all of this yourself.
Not every computer is the same. WebSLM checks WebGPU, memory and storage, suggests a model that fits and flags any that may be slow.
Small models aren't the largest cloud models. They make mistakes and know less, so check the facts they give you. For a large share of everyday writing, they're the right tool.
How it works
The app checks your device and suggests one of 11 light open models, from about 150 MB to 1 GB. Each shows its size and licence.
Your browser fetches the model file from Hugging Face and keeps it in its cache. Next time it loads from disk in seconds.
llama.cpp runs the model on your device, using WebGPU when available and your processor otherwise.
Models
Open models small enough to run in a browser tab on a laptop or desktop. Before a model is offered, we check on Hugging Face that the file exists, needs no account, and runs in this engine. WebSLM doesn't bundle or modify them.
Smallest first. Scroll for more. The app also offers 10 larger models for powerful computers.
Privacy
The full details are in the privacy policy.
Before you start
WebSLM is free open-source software under the Apache-2.0 licence, provided "as is", without warranties of any kind. There is no paid support or uptime guarantee.
Small models can produce confident text that is incorrect, outdated, incomplete or biased. Review everything before you rely on it or share it.
Don't use WebSLM in place of medical, legal, financial or other professional advice, or to make decisions that significantly affect people.
The models are made by Google, Microsoft, Meta, IBM, Mistral AI, Alibaba Qwen, Liquid AI, Hugging Face and others, not by WebSLM. You are responsible for following each model's licence and usage policy. For example, Liquid AI's LFM licence is free only for organizations under US$10M in annual revenue.
A model takes about 150 MB to 1 GB of storage and more memory while it runs. Generating text keeps your processor or graphics card busy and drains battery faster. Use Wi-Fi for the first download if your data plan is limited. You can remove models at any time in the app.
Your chat history stays in this browser only, until you delete chats or clear site data. You can export chats to a file to back them up or move them. On a shared computer, delete your chats when you're done.
Loading this site contacts GitHub Pages, downloading a model contacts Hugging Face, and signing in uses Supabase, which stores your email and your model downloads. Anonymous visit counts go to GoatCounter. Your prompts are not sent to any of them.
WebSLM runs models with llama.cpp through wllama, reads PDFs with pdf.js and Word files with mammoth. Credits and licences are in the NOTICE file.
FAQ
Your prompts and the model's answers are computed in your browser and never sent to us, because we don't run a chat server at all. What we do keep: your email address (for sign-in) and a record of which models you download. Loading the site uses GitHub Pages, and models come from Hugging Face.
So we can see how many people use WebSLM and which models they download, which tells us where to focus. Sign-in is free and uses a one-time link sent to your email, so there's no password. Models you've already downloaded keep working without signing in, and your chats are never part of it.
Yes: PDFs, Word documents (.docx) and text files such as .txt, .md, .csv, .json, .html, subtitles, logs and code. Click the paperclip, drag a file onto the message box, or paste it. Your browser reads the file; it's never uploaded. Scanned PDFs (images of text) can't be read yet. Small models can only read so much at once, so for long files the app tells you when only the first part fits, and you can raise Memory in Settings.
Nothing. WebSLM is free and open source under the Apache-2.0 licence.
A recent desktop browser on a laptop or desktop. Browsers with WebGPU (such as Chrome and Edge) are fastest; without it, WebSLM uses your CPU through WebAssembly. Light models are about 150 MB to 1 GB to download. Larger ones (1 to 2 GB) are offered for powerful computers only.
The app checks your device first: whether WebGPU is available, roughly how much memory your browser reports, your CPU threads and your free storage. It suggests a model and marks each one as a good fit, possibly slow, or too big for your free storage. These checks happen on your device. As a rule of thumb, start with Liquid LFM2.5 1.2B on a recent laptop and with LFM2.5 350M or SmolLM2 360M on older machines.
The current model is unloaded from memory but stays downloaded, and the new one loads (from your browser if you've used it before). Your conversation carries on: the new model sees the chat so far, and each answer is labelled with the model that wrote it. You can't switch while an answer is being written; press Stop first.
Open the model picker at the top of the app: each downloaded model has a Remove link, and Settings has Remove all downloaded models. Removing a model frees the space right away and keeps your chats.
No, and it isn't trying to be. Small models are best at rewriting, summarizing, translating and drafting. They can be wrong, so check facts.
The WebSLM code is Apache-2.0. Each model has its own licence, shown next to it in the app. For example, Liquid AI's LFM models are free for organizations under US$10M in annual revenue, while Granite, Qwen3 and SmolLM2 are Apache-2.0. Check the licence and your organization's policies.
Yes. After your first visit and model download, the app and the model are cached by your browser, so you can keep chatting without a connection.
Something else? Ask on GitHub or email webslmteam@gmail.com
WebSLM
Free and open source. Runs in Chrome, Edge and other modern browsers on your laptop or desktop.