Which AI models are there?
Archinode has nine image models, ordered from most to least realistic. All of them are in Canvas AI and in the Nodes Generator; AI Chat uses four of them. The cost depends on the resolution:
| Model | What it's best at | Tokens per render (1K · 2K · 4K) | Plans |
|---|---|---|---|
| Archinode AI | Affordable, with a large fixed-size output. A single resolution. | 20 · — · — | All, including the trial |
| NanoBanana 2.1 | The newest NanoBanana: faster than PRO, with better visual design and consistency. 1K, 2K or 4K. | 20 · 30 · 50 | Paid (from Basic) |
| NanoBanana PRO | The most realistic, and the best at understanding detailed prompts. | 45 | From Standard |
| NanoBanana Turbo | Good realism and fast. You have to spell out every detail you expect. | 20 · 20 · — | Paid (from Basic) |
| NB Lite | The most affordable NanoBanana. Fixed 1K output, ideal for quick tests. | 15 · — · — | Paid (from Basic) |
| GPT Image PRO | PRO's 2.6 version: the most detail, for final renders and presentation pieces. Offers Complexity and Transparent background. | 10 · 20 · 20 | From Standard |
| GPT Image Turbo | Turbo's 2.6 version: fast and versatile, great for iterating ideas and variations. Offers Complexity and Transparent background. | 10 · 20 · 20 | From Standard |
| Flux 3 | Edits one element and keeps the rest. The newest of the Flux family; goes up to 2K. | 20 · 70 · — | Paid (from Basic) |
| Krea 3 | Edits with up to eight images and keeps the rest of the frame. Goes up to 2K. | 20 · 60 · — | Paid (from Basic) |
| Flux 3 Klein | Fast and very affordable, with or without a base image. Combines up to ten images and goes up to 4K. | 10 · 10 · 20 | Paid (from Basic) |
| Grok Imagine | Fast and affordable, with two quality levels. Takes up to five images. | 30 · 30 · — | Paid (from Basic) |
| Seedream 5 | Realistic, good detail and good editing. Expensive. | 50 · 50 · — | From Standard |
| Seedream 4 | Realistic with good results. Cheap. | 10 · 20 · 40 | Paid (from Basic) |
| Flux 2 | Realistic with good results. Cheap. | 10 · 20 · 40 | Paid (from Basic) |
| Qwen 3 Edit | The cheapest edit. Takes up to three images. | 20 · 30 · 60 | Paid (from Basic) |
| Boogu Edit | Edits while preserving the source render as much as possible. Uses only the base image. | 20 · 50 · 120 | Paid (from Basic) |
A “—” means the model doesn't reach that resolution: NanoBanana Turbo and Seedream 5 stop at 2K. Each variant is a render and is charged separately (except on the Canvas AI engine, where Archinode AI and NanoBanana PRO variants after the first cost 10 and 25), and GPT Image and Seedream 5 add 5 tokens for each extra image you connect. On the free trial everything renders at 1K; 2K starts at Basic and 4K at Standard.
Seedream 5, Qwen 3 Edit, Boogu Edit, Flux 3, Grok Imagine, Archinode AI and NB Lite are editing models: they need a base image and change one part while leaving the rest as it is. The others can also generate from scratch (in Nodes, from text alone).
GPT Image PRO, GPT Image Turbo and Grok Imagine have a quality selector (Fast, Standard and High; Grok only Fast and Standard). More quality gives more detail and costs more; if you don't choose, Standard is used, which is the price in the table. Muse and NB Lite deliver a fixed size, so they only offer 1K.
If a model shows as “Unavailable” in the menu, it's paused for the moment: pick another one.