WebLLM · 100% in-browser
Free LLM chat online, no login required
Ask questions, write, or chat with a document using open models on your own device. Choose a model, approve its download, and start a conversation. Your chat stays local unless you enable an external search provider.
Start small: SmolLM2 360M downloads ~220 MB and requires WebGPU. Device compatibility is shown before download.
Why it's different
Free means free — and private means private
-
No account required
Open the studio, choose a model, and approve its download before your first message.
-
Local inference
Your device generates the answer. Chat content stays local unless you choose an external search provider.
-
No message limits
No daily cap, no throttling, no queue. Chat as long as your browser tab is open.
-
No subscription
The core chat is free. Model downloads use your connection and browser storage; inference uses your device.
Before your first answer
Check your device and model download
Use a current browser and check the device compatibility label in the chat's model selector. WebLLM models need WebGPU. If it is unavailable, LFM2.5 offers a slower CPU fallback. On phones, start small and keep the tab open while loading.
The starter model, SmolLM2 360M, downloads ~220 MB; larger models need more storage and memory. You approve downloads before they begin. Cached files can be reused, but offline use requires all model and runtime files to be available. Test before disconnecting. Compare chat models and device requirements before choosing.
More than a blank prompt
Local documents and voice, with search when you choose
-
PDF knowledge (RAG)
Drop in a PDF and the chat answers from it — parsed and embedded locally, page by page.
-
Optional web search
Bring your own API key to let the assistant pull in live web results when you want fresh answers.
-
Voice mode
Speak your prompts and hear replies with local Whisper transcription and Kokoro TTS.
Questions
Frequently asked questions
-
Do I need to sign in or create an account?
No. There is no sign-in, no email form, and no account to create — the chat is ready the moment the model finishes loading in your browser.
-
Is the chat really unlimited?
Yes. Because the model runs on your own device, there is no daily cap, no throttling, and no queue — you can chat for as long as the tab stays open. See how this compares to ChatGPT.
-
Which LLMs can I chat with?
Start with SmolLM2 360M (~220 MB) on a WebGPU-capable browser, or choose a larger Qwen model if your device has enough memory. LFM2.5 also supports a slower CPU fallback. The model selector shows device compatibility.
-
Does it work offline?
Offline use depends on every required model and runtime file being cached. Finish setup and test the chat before disconnecting. Browsers may evict cached files when storage is low.
-
Will this work on my phone?
It depends on browser support and available memory. Check the compatibility label in the model selector, start with a small model, and keep the tab open during setup. LFM2.5 can use a slower CPU fallback when WebGPU is unavailable.
-
Why does the first reply take longer?
Before the first reply, your chosen model downloads and initializes on your device. Download size appears before approval, and progress appears in the chat. Later loads can reuse cached files. Download and generation speeds depend on your connection and hardware.
Ready for your first local conversation?
Pick a model, let it download once, and chat as much as you like. No account, no cap, and local inference.