Local AI, with the model you choose.
Every suggestion comes from a model on your Mac, run by Typeahead’s own engine, built to be fast and light. Gemma and Qwen are built in, or bring your own.
Local, and it shows.
Quick as you type, the model you choose, and it works offline after setup.
Instant.
Suggestions keep pace as you type.
Your own model.
Point it at any GGUF file.
Wi-Fi off? Still there.
Works offline after setup.
Or bring your own model.
Typeahead recommends the built-in model that suits your Mac. Prefer another? Point it at a GGUF model file, like Llama or Mistral.
In detail.
The model, your Mac, and what it needs.
The built-in models.
Gemma and Qwen come in Typeahead’s catalog. During setup it recommends the one that suits your Mac’s memory: a Gemma model by default, a smaller Qwen model for lighter Macs, and a larger Gemma model for Apple Silicon Macs with 24 GB of memory or more. Switch in Settings › Model whenever you like. Updates never change the model you picked.
Bring your own model.
Under Settings › Model › Custom Model, point Typeahead at your own GGUF model file instead of one from the catalog. Instruct and chat models work best, from families like Llama and Mistral. Every model is included on every plan.
Checked downloads.
Built-in models download once, from a pinned version, and each file is checked with SHA-256 before Typeahead uses it. An interrupted download picks up where it left off.
Its own engine, built to be light.
Typeahead runs your model on its own engine, built to be fast and energy-efficient. Your keystrokes are handled first. After 15 minutes without typing, the model sleeps and frees its memory, then wakes the moment you type again. If your Mac runs short of memory, Typeahead pauses and its suggestions come back on their own. Nothing waits on a server, so its AI autocomplete keeps pace as you type.
What you need.
macOS 14 Sonoma or later, at least 8 GB of memory, and about 4 GB of free disk space. It asks for Accessibility permission only, not Screen Recording. It works in the Mac apps you already use, and the compatibility page lists them.
Common questions
It’s autocomplete whose AI model runs on your own computer, not on a company’s server. Typeahead downloads a model to your Mac once, during setup, and its own engine makes every suggestion there as you type in your Mac apps. You choose the model: Gemma and Qwen are built in, or bring your own GGUF file.
Start with the one Typeahead recommends for your Mac. The default Gemma model is a strong all-rounder and the best pick for languages other than English. The smaller Qwen model is lighter and English-first. On an Apple Silicon Mac with 24 GB of memory or more, a larger Gemma model gives higher-quality suggestions. Switch any time in Settings › Model.
Yes. Under Settings › Model › Custom Model, point Typeahead at a GGUF model file instead of one from the catalog. Instruct and chat models work best, from families like Llama and Mistral. Every model is included on every plan.
It’s built not to. Typeahead runs on its own engine, built to be fast and light: your keystrokes are handled first, the model sleeps after 15 minutes without typing and frees its memory, and if your Mac runs low on memory, suggestions pause and come back by themselves.
macOS 14 Sonoma or later, at least 8 GB of memory, and about 4 GB of free disk space. Typeahead asks for the Accessibility permission only; it doesn’t need Screen Recording.
Not for autocomplete. There’s no round trip to a server, so suggestions keep pace as you type. How quick it feels depends on your Mac and the model you pick.
