Turn on airplane mode, disconnect from Wi-Fi, and most software you rely on daily starts breaking in small ways — sync stalls, spell-check downgrades, "AI" features go gray. Typeahead doesn't, because it was never built to depend on a connection for the thing it actually does: suggesting the rest of the sentence you're typing.
Why This Even Needs Proving
Most writing-assistance tools on the market today route your keystrokes to a cloud API to generate suggestions. That's a reasonable architecture for a large model that doesn't fit on a laptop — but it means the feature simply stops working the moment there's no connection, and it means whatever you're typing is transmitted to a server somewhere in between. Typeahead takes the opposite approach: the model that generates suggestions lives on your machine.
What's Actually Running Locally
A local LLM via llama.cpp
Typeahead runs Qwen 3 through llama.cpp directly on your PC. Once the model is downloaded, generating a suggestion is a local inference call — no request leaves your machine.
Snippets and dictionary autocomplete
Custom autocorrect rules and dictionary-based autocomplete work even before a model is downloaded at all — they don't depend on the LLM or a connection in any form.
No background sync to interrupt
There's no account, no cloud settings sync, and no telemetry on what you type. There's nothing running in the background that airplane mode would need to pause.
The only two moments Typeahead ever needs a network connection are downloading a model the first time and checking for app updates. Neither of those happens mid-sentence. Once a model is on disk, suggestion generation is entirely offline from that point forward.
Where This Actually Shows Up
The scenario is simple: laptop in airplane mode, no Wi-Fi, writing in whatever app you'd normally use. Typeahead keeps suggesting the rest of your sentence in Notepad and Notepad++, drafting a document in Word, catching up on email offline in Outlook or Thunderbird, or writing in a chat pane in VS Code. Suggestions appear as gray "ghost text" ahead of the cursor exactly as they would with a full signal, and Tab still accepts them.
The same is true for apps that are normally network-dependent themselves, like Slack, WhatsApp, Telegram, Chrome, Edge, and Firefox. Those apps may show you a "no connection" banner of their own for their own sync — but the text field you're typing into still gets Typeahead's suggestions, because the suggestion engine doesn't route through those apps' network layer at all; it watches the caret directly at the OS level.
Why Offline-First Was the Design, Not an Afterthought
This isn't a fallback mode bolted on for edge cases — it's the default architecture. Typed text never leaves the device, there's no text telemetry, and there's no account to be signed out of when you lose signal. A tool that suggests the private, half-formed sentence you're in the middle of writing has a stronger reason than most to keep that sentence local, connection or not.
Frequently Asked Questions
Does Typeahead really work with no internet connection at all?
What still works even before I download a model?
Which apps does offline suggestion support cover?
Does it work in other text editors too?
Is anything I type stored or sent once I reconnect to Wi-Fi?
Get Typeahead
One-time purchase. No subscription. Works on Windows 10 and Windows 11.
Get it on Microsoft Store