AI chat
Several language models in one service. Pick a model for the task, attach a photo or a document and, if you like, ask for an image — it appears right in the conversation. The chat works in the web studio.
Models and prices
| Model | Good at | Attachments | Images in chat | Answer, approx. |
|---|---|---|---|---|
| Claude Opus 5.5 | The smartest: hard tasks, code, long texts | images, PDF, text files (TXT, MD, CSV, JSON) · up to 5 files, up to 5 MB each | yes | ≈ 0.3 t |
| Claude Sonnet 5.5 | Smart and fast: writing, code, analysis | images, PDF, text files (TXT, MD, CSV, JSON) · up to 5 files, up to 5 MB each | yes | ≈ 0.2 t |
| GPT-6 Sol | OpenAI's strong all-round model | images, PDF, text files (TXT, MD, CSV, JSON) · up to 5 files, up to 10 MB each | yes | ≈ 0.1 t |
| GPT-6 Luna | Fast and very cheap | images, PDF, text files (TXT, MD, CSV, JSON) · up to 5 files, up to 10 MB each | yes | ≈ 0.1 t |
| Gemini 3.8 Flash | Google's fast model | images, PDF, text files (TXT, MD, CSV, JSON) · up to 5 files, up to 10 MB each | yes | ≈ 0.1 t |
| Grok 4.7 | xAI's model: reasoning and code | images, text files (TXT, MD, CSV, JSON) · up to 5 files, up to 10 MB each | yes | ≈ 0.1 t |
| DeepSeek V4.1 Flash | Fast and cheap, text only | text files (TXT, MD, CSV, JSON) · up to 3 files, up to 2 MB each | no | ≈ 0.1 t |
How the price of an answer works
An answer is billed by what it actually used: how much text the model read — your message, the chat history and attachments — and how much it wrote back. A short question costs less, going through a long document costs more.
The table shows a typical answer: about 2,000 units of text in and 800 out (that is how models count text, roughly a page and a half of question and a page of answer). The exact amount is shown under every answer in the chat; the minimum is 0.1 token.
Images right in the conversation
Models marked “yes” under “Images in chat” start a generation themselves when you ask them to draw something: they write a detailed prompt and hand it to an image model. The finished image appears in the conversation, where you can open and download it.
The image model is chosen in the chat settings. The image is billed at that model’s price, like any generation in the studio.
Chats and history
Every conversation is kept as a separate chat in the list on the left: rename it, delete it or come back to it later. The model sees the history of the current chat, so you can refine without repeating yourself. When a conversation gets too long for the model, the studio suggests starting a new chat.