Kayts has five models in one chat, and you can pick a different one for each question from the model menu. They are not five versions of the same thing: each is built for a kind of work. This is what each one is for.
The five, in short
- Kayts Mihr (every plan): fast and free. Quick answers, drafts and rewrites, summaries of photos, PDFs and spreadsheets, everyday questions in your own language.
- Kayts Anahit (from Plus; a daily taste on Free): longer writing and careful editing, step-by-step reasoning and analysis, research with several sources and citations.
- Kayts Siran (from Pro): detailed analysis and structured reports, writing, explaining and fixing code, multi-step problems that need care.
- Kayts Nareia (from Pro): very long documents and whole books, projects that span many conversations, research over many sources at once. It reads text and files, not images yet.
- Kayts Monte (Max): the hardest questions and the most careful answers, long and complex documents with images, work where a mistake is expensive.
How much each one keeps in view
A model's context window is how much it can hold at once: your conversation, the files you added and the sources from Research. Past it, the oldest parts no longer fit. Three of the five hold about four times more than the other two.
Each model's page on kayts.ai (Products → the model), checked 21 October 2026.
Show the data as a table
| Value | |
|---|---|
| Kayts Mihr | 262K |
| Kayts Anahit | 262K |
| Kayts Siran | 1M |
| Kayts Nareia | 1M |
| Kayts Monte | 1M |
A token is about three quarters of an English word. 262 thousand tokens is roughly 190,000 words, about two novels; a million is roughly 700,000 words, a small shelf of books. For an article or a spreadsheet either is plenty; the difference matters when you work with whole books, a long thesis or many sources together.
A simple way to choose
- Start with Mihr. If the answer is good, you are done; it has no daily counter, only fair use.
- Switch to Anahit when the writing needs more care or the question has several steps.
- Take Siran for code and long analysis, Nareia when the material is very long, and Monte when the answer has to be as good as it can be.
Every model can think before it answers when you turn on Thinking, and every one can look up papers on its own. Each plan has a daily allowance for the stronger models; when it runs out, the chat continues on Kayts Mihr and tells you. More about each model: all models; their published test scores: our article on benchmarks.
