Leonardo AI Phoenix and Models: Which to Use and When

Leonardo gives you a choice of models for every generation, and the choice matters more than any other setting. Phoenix is Leonardo’s own model and the sensible default for most work; the Flux variants and older models each have jobs they do better. This guide explains what to pick, how to write for it and which settings to touch first.

What Phoenix is for

Phoenix is the in-house model Leonardo develops itself. In practice it is the one to reach for when you need the image to follow a detailed prompt: several named objects in specific positions, a particular colour scheme, readable short text on a sign or label, or a composition described in full sentences. It handles long, descriptive prompts well and tends to keep the elements you asked for rather than dropping one.

Phoenix also pairs with Leonardo’s Contrast setting and Prompt Enhance. Contrast controls how punchy the tonal range is; a lower value gives a softer, flatter image that suits illustration and product work, a higher value gives deep shadows and bright highlights that suit dramatic concept art. Prompt Enhance rewrites a short prompt into a longer, more detailed one before generation. It is useful when you have only a few words; switch it off when you have written a careful prompt and want it followed exactly.

The other models and when to switch

Leonardo also offers versions of Flux, a model family known for natural-looking photography and clean rendering of hands and faces, alongside a set of older Leonardo models and fine-tuned community models. Exact names and availability change, so check the model menu in the current interface, but the decision logic stays the same:

• Photoreal people, products and interiors: try a Flux variant. Compare with Phoenix on the same prompt and seed before committing.
• Dense, specific prompts and legible text: Phoenix.
• A particular established look (anime, 3D render, flat vector): a fine-tuned model built for that look, or Phoenix plus an Element (see Elements and Image Guidance).
• Fast iteration: whichever model offers a speed-optimised mode in your plan. Use it for composition tests, then re-run the keeper on the quality mode.

Each model has its own token cost per image, and some are more expensive than others. Leonardo runs on a token-based plan, so check the current pricing page rather than relying on numbers from a forum post.

Writing a prompt Leonardo follows

Whichever model you pick, the same structure works: subject, then what it is doing, then setting, then light, then style and medium. Leonardo reads natural sentences well, so write "a ceramic teapot with a cracked blue glaze sits on a sunlit wooden table, steam rising, shallow depth of field, editorial food photography" rather than a comma list of keywords.

Three Leonardo-specific points:

1. Preset Styles (Cinematic, Illustration, Dynamic and so on) apply a hidden style layer on top of your prompt. They are a quick way to set a look, but they can override your own style words. If a prompt is ignoring your instructions, set the preset to none and describe the style yourself.
2. The negative prompt is available on most models. Use it for recurring problems ("text, watermark, extra fingers, blurry") rather than for everything you can think of; long negatives dilute the main prompt.
3. Prompt Enhance changes your words. Read the expanded prompt when you can see it, so you know what the model actually received. Keep the version you liked.

Settings at a glance

Names and positions move between interface updates; these are the controls worth finding.

A simple model test

When you are not sure which model suits a job, run this before spending tokens on dozens of variations:

1. Write one careful prompt with Prompt Enhance off and no Preset Style.
2. Generate two images with Phoenix and note the seed of the better one.
3. Run the same prompt and seed on your Flux variant and on any fine-tuned model you are considering.
4. Compare for adherence (is everything you asked for present?), realism and the specific thing the job needs, such as readable text or convincing skin.
5. Pick the model, then turn on Prompt Enhance or a Preset Style only if the result still needs a push.

Record the model, seed and settings next to the image you keep. The Leonardo history view shows them, but a note of your own is easier to search in six months.

Try it

Paste this into Phoenix with Prompt Enhance off, Contrast medium and no Preset Style:

A small independent bookshop at dusk seen from across a wet cobbled street, warm light in the window, a hand-painted sign above the door reading "OPEN LATE", a bicycle leaning on the wall, soft rain, 35mm photograph, muted teal and amber palette

Change the sign text first to see how reliably Phoenix renders short words, then swap "35mm photograph" for "gouache illustration" to see how much the medium phrase alone moves the image before you reach for a Preset Style.