Getting started
No sign-up, no configuration. Open the chat, type, and press Enter. This page covers the first conversation and what each part of the screen does.
Your first message
Go to chathaiku.com/chat. The box at the bottom of the screen is the composer. Type a question and press Enter to send it. Shift + Enter adds a new line instead of sending, which is useful for pasting several paragraphs or writing a multi-part question.
A reply begins arriving within a second or two. While it is being written, the text shimmers and the model's avatar picks up a spinning ring — both mean generation is still in progress and the reply is not yet complete.
The layout
The interface differs between desktop and mobile, because the space available is different. Nothing is missing on mobile — the same features are reached from a menu rather than a permanent sidebar.
On desktop
| Where | What |
|---|---|
| Far left rail | New chat, search, settings, and a link to this support site. |
| Top left | Sidebar toggle, and a search button for finding past conversations. |
| Left sidebar | New chat button, your saved conversations, and Settings at the bottom. |
| Top centre | Model picker. Shows the active model; click to switch. |
| Top right | Outline toggle — a jump list of every turn in the current conversation. |
| Bottom | Composer, with response controls behind the three-dot button. |
On mobile
The sidebar is hidden. Tap the chevron next to the ChatHaiku title to open a menu containing New chat, Sessions, Settings, and quick toggles for theme and voice.
Pasting long text
Pasting something substantial — roughly six lines or more — lifts it out of the text box into a labelled chip above the composer instead of filling the box with hundreds of lines you then have to scroll through.
The pasted content is sent with your message, ahead of whatever you type, so the usual pattern — paste a block, then ask a question about it — reads the way you would expect. Several pastes can be attached at once, and anything still attached counts toward the message limit below.
Message length
Messages are limited by word count, and the limit depends on the model:
| Model | Limit per message |
|---|---|
| Epos models | 2,048 words |
| Haiku H2 | 200 words |
H2's limit is tighter because it has a 1,024-token context window — a long prompt would crowd out the space it needs for its own reply. Going over the limit does not silently truncate: the message is refused with a note saying the limit and your actual word count, so you can shorten it and try again. Pasted attachments count toward the total.
Keyboard shortcuts
| Keys | Action |
|---|---|
| Enter | Send the message |
| Shift + Enter | New line without sending |
| Ctrl / ⌘ + K | Search your conversations (desktop) |
| ↑ ↓ | Move through search results or menu items |
| Esc | Close any open panel, menu, or dialog |
What to expect from the models
ChatHaiku serves small models. The largest is 1.5 billion parameters; the smallest is 217 million. For comparison, the frontier assistants most people have used are two to three orders of magnitude larger. That difference is real and you will notice it.
- They are fast. Replies start almost immediately and run on a single consumer graphics card.
- They get facts wrong. Small models produce fluent, confident text whether or not the underlying knowledge is there. Treat factual claims as a starting point.
- They are better at some things than others. Clear explanation and conversation are strong. Structured planning — itineraries, budgets, schedules — is measurably weak.
- They have a knowledge cutoff. Epos 2.1 can look some things up (see sources), but coverage is narrow.
The disclaimer under the composer is not boilerplate. Do not rely on any model here for medical, legal, financial, or safety-critical decisions. None of them have been evaluated for those uses.