Which AI image model follows a long, detailed prompt best?

No model drew every detail of this prompt, and one image per model is not a ranking. All 15 drew the main scene. The split came on small or unusual details: four models drew the tiny figure walking upside down, Nano Banana 2, Nano Banana 2 Lite, Flux 2 Flex and Grok Imagine 2.

Which details do AI image models skip in a long prompt?

In this test, the missed details were unusual, tiny or optical: water flowing upward, a small figure walking upside down in the distance, and distorted copies of the woman in the glass. Common things, such as the woman, the moon and the city lights, appeared in every image.

Can an AI image model draw a waterfall flowing upward?

None of the 15 models did in this test. Thirteen drew waterfalls, and every one of them falls down. The two Grok models drew no waterfalls at all.

How much does an image cost on these models?

At today's prices and the settings we used, an image costs from 4 credits with Flux 2 Klein to 77 credits with Flux 2 Max. Every model's price is in its caption on this page.

Can I compare image models side by side in Nodaro?

Yes. Connect one Text node with your prompt to several Generate Image nodes, choose a different model in each node, and run the workflow. Each node shows its price on its Run button before it runs.

AI image model comparison: 15 models, one long, detailed prompt

We gave 15 AI image models one long prompt with 15 hard details. See which details each model drew, from the crystal city to the figure walking upside down.

We gave 15 AI image models the same long prompt. It describes a woman in a white dress at the edge of a city inside a giant crystal, under a red moon, with whales in the sky and a tiny figure walking upside down. Every model drew the woman, the moon and the city lights. The models split on the hard details: only four drew the figure walking upside down, and none drew water flowing upward.

Models: 15
Prompt: 1
Details checked: 15
Aspect ratio: 16:9
Upside-down figure: 4 of 15
Upward waterfalls: 0 of 15

The test

Every model got the same prompt and a 16:9 aspect ratio, and we asked each model for 2K. Models without a 2K setting made their usual size, so the images range from 1280×720 to 2752×1536. Each model made one image, its first take. We ran the test in September 2026.

This is the prompt, word for word:

The image prompt

A lone woman in an elegant flowing white couture dress stands at the edge of an impossible city built inside a colossal transparent crystal suspended above a dark ocean. Inside the crystal, an entire surreal metropolis stretches into the distance: monumental glass towers bend and curve like liquid, upside-down buildings hang from the sky, enormous translucent staircases climb into clouds, and tiny waterfalls flow upward between floating islands. A gigantic crimson-red moon fills almost half the sky behind the crystal, partially obscured by dramatic storm clouds. Thousands of tiny warm lights glow throughout the city, creating an extraordinary sense of scale. The woman stands on a narrow mirror-like black platform extending over the void. Her dress is caught by a strong wind, with long pieces of fabric flowing dramatically behind her. Below the transparent floor, enormous luminous blue whales appear to be swimming slowly through the night sky instead of water. Rain falls outside the crystal but stops perfectly at its surface. Individual droplets cover the exterior, refracting and distorting fragments of the city behind them. The crystal creates complex optical phenomena: refraction, internal reflections, chromatic dispersion, subtle rainbow caustics, distorted duplicates of the woman, and fractured reflections of the red moon. Far in the distance, a tiny human figure walks upside down across the underside of one of the floating islands. The image should feel enormous, mysterious, beautiful, physically impossible, and strangely believable. Extreme contrast of scale. Deep spatial layering from foreground to extreme distance. Intricate transparent materials. Realistic human proportions. Rich micro-detail. Atmospheric depth. Volumetric light passing through crystal and mist. Complex reflections and refractions. Physically convincing surfaces despite the impossible environment. The composition should immediately make the viewer stop and examine the image to understand what they are seeing.

We split the prompt into 15 details and checked every image for each one.

Which details the models drew

#	Detail in the prompt	Drawn	Partly	Missing
1	A woman in a white couture dress, caught by the wind	15	0	0
2	A narrow, mirror-like black platform	14	1	0
3	A colossal crystal above a dark ocean	10	4	1
4	The city inside the crystal	13	1	1
5	Glass towers that bend like liquid	8	3	4
6	Upside-down buildings hanging from the sky	7	4	4
7	Translucent staircases climbing into clouds	11	1	3
8	Waterfalls flowing upward	0	13	2
9	A giant crimson moon behind the crystal	15	0	0
10	Thousands of tiny warm lights	15	0	0
11	Luminous blue whales below the floor	12	3	0
12	Rain that stops at the crystal's surface	9	3	3
13	Rainbow light in the crystal	6	4	5
14	Distorted copies of the woman or the moon	5	6	4
15	A tiny figure walking upside down under an island	4	1	10

In row 8, Partly means that the image has waterfalls, but they fall down, as waterfalls usually do.

What we saw

The scene

Every model drew the main scene.

All 15 drew the woman in the wind, the giant red moon and the city lights. 14 put her on a narrow black platform, and 13 put the city inside the crystal.

Water

No model drew water flowing upward.

13 images have waterfalls, and every one of them falls down. The two Grok models drew no waterfalls at all.

The figure

Four models drew the figure walking upside down.

Nano Banana 2, Nano Banana 2 Lite, Flux 2 Flex and Grok Imagine 2 drew it. Flux 2 Pro hung a figure from the edge of an island instead.

Optics

The optical effects split the models.

Six drew rainbow light in the crystal. Five drew distorted copies. Seedream 5 Pro and Grok Imagine 2 drew a ghost of the woman, Nano Banana 2 Lite and Flux 2 Pro a second moon, and Grok Imagine broken reflections of the moon.

Rain

Nine kept the rain outside the glass.

In these images the rain falls outside the crystal or beads on its surface, and the city inside stays dry. Flux 2 Klein, Flux Kontext Pro and Flux Kontext Max let the rain fall everywhere.

The crystal

Ten models floated the crystal above the sea.

Seedream 5 Lite set it on the water, and Flux Kontext Max made it a tower that rises from the sea. Flux 2 Klein drew a glass floor instead, with the city on an island.

The full checklist

The numbers are the details in the table above. ✓ means drawn, ~ means partly drawn, and ✗ means missing.

Model	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15
Flux 2 Flex	✓	✓	✓	✓	~	✓	~	~	✓	✓	✓	✓	✗	~	✓
Flux 2 Klein (Open)	✓	~	✗	✗	✓	✓	✓	~	✓	✓	✓	✗	~	~	✗
Flux 2 Max (Safety Tolerance)	✓	✓	✓	✓	✗	~	✓	~	✓	✓	~	✓	✓	✗	✗
Flux 2 Pro	✓	✓	✓	✓	✗	✗	✓	~	✓	✓	~	✓	✓	✓	~
Flux 2 Pro (Safety Tolerance)	✓	✓	✓	✓	✓	✗	✓	~	✓	✓	✓	✓	✓	✗	✗
Flux Kontext Max	✓	✓	~	~	✗	✗	✗	~	✓	✓	✓	✗	✗	~	✗
Flux Kontext Pro	✓	✓	✓	✓	✗	✗	✗	~	✓	✓	~	✗	✗	~	✗
GPT Image 2	✓	✓	~	✓	~	✓	✓	~	✓	✓	✓	✓	✓	✗	✗
Grok Imagine	✓	✓	✓	✓	✓	✓	✗	✗	✓	✓	✓	~	✗	✓	✗
Grok Imagine 2	✓	✓	✓	✓	✓	~	✓	✗	✓	✓	✓	~	✓	✓	✓
Nano Banana 2	✓	✓	✓	✓	✓	✓	✓	~	✓	✓	✓	✓	✓	~	✓
Nano Banana 2 Lite	✓	✓	✓	✓	✓	✓	✓	~	✓	✓	✓	✓	~	✓	✓
Seedream 5 Lite	✓	✓	~	✓	✓	✓	✓	~	✓	✓	✓	~	✗	✗	✗
Seedream 5 Pro	✓	✓	~	✓	~	~	✓	~	✓	✓	✓	✓	~	✓	✗
Wan 2.7 Pro	✓	✓	✓	✓	✓	~	✓	~	✓	✓	✓	✓	~	~	✗

The models are in alphabetical order, not ranked, and the checklist has no total on purpose. The details a prompt depends on differ from one project to the next.

Every image

Each image below is a web copy. The link under each image opens the full-size original.

Alibaba

Wan 2.7 Pro: a woman in a white dress faces the camera on a black glass platform inside a glass bowl, with curved towers, staircases and a giant red moon behind her — Wan 2.7 Pro2560×1440 · 66 credits
The woman faces the camera, inside a glass bowl in front of a huge red moon. The tiny figure on the right stands upright.
Full-size original

Black Forest Labs

Flux 2 Flex: a woman in a white dress on a black platform beside a tall crystal shard holding a city, with figures hanging upside down under floating islands — Flux 2 Flex1920×1088 · 66 credits
Two figures hang upside down under the floating islands at the top. There is no rainbow light in the crystal.
Full-size original

Flux 2 Klein: a woman in a white dress on a glass floor over the sea, with curved towers, upside-down buildings and whales under the glass, in heavy rain — Flux 2 Klein (Open)1360×768 · 4 credits
The city stands on an island around a wide glass floor, not inside a crystal. Rain falls everywhere, and the whales swim under the glass floor.
Full-size original

Flux 2 Max: a woman in a white dress on a black path in front of a giant cut crystal holding a city, under a red moon, with grey whales and rain — Flux 2 Max (Safety Tolerance)1888×1072 · 77 credits
A giant cut crystal over a stormy sea, with rainbow facets and droplets. The towers are straight, the whales are grey, and a figure falls through the air outside.
Full-size original

Flux 2 Pro: a woman in a white dress walks toward a diamond-shaped crystal with a city, a spiral staircase and a second red moon inside it — Flux 2 Pro1920×1088 · 22 credits
A second, refracted moon appears inside the crystal. A figure hangs from the edge of an island, not under it.
Full-size original

Flux 2 Pro (Safety Tolerance): a woman in a white dress on a platform watches a diamond crystal holding a city, a spiral staircase and a twisted tower — Flux 2 Pro (Safety Tolerance)1888×1072 · 26 credits
A spiral staircase and a twisted tower fill the crystal. The woman watches from a separate platform, and no buildings hang upside down.
Full-size original

Flux Kontext Max: a woman in a white dress on a wet ledge facing a crystal tower wrapped in a glowing city, with bright blue whales and heavy rain — Flux Kontext Max1392×752 · 28 credits
The city wraps around a crystal tower that rises from the sea, and the rain falls in front of it. The whales glow brightly.
Full-size original

Flux Kontext Pro: a woman in a white dress on a black path toward a ring of crystal that holds a city and a giant red moon, in heavy rain — Flux Kontext Pro1392×752 · 15 credits
A ring of crystal holds the city and the moon. Heavy rain falls everywhere, and there are no staircases.
Full-size original

ByteDance

Seedream 5 Lite: a woman in a white dress on a black platform facing a crystal dome on a stormy sea, with a city, a red moon and many glowing whales — Seedream 5 Lite2560×1440 · 16 credits
The crystal dome sits on the stormy sea instead of floating above it. There is no rainbow light and no upside-down figure.
Full-size original

Seedream 5 Pro: a woman in a white dress on a black platform beside a glass bubble holding a city, with two ghost copies of her reflected in the glass — Seedream 5 Pro1376×768 · 20 credits
Two ghost copies of the woman appear in the glass, and droplets bead on its surface. The tiny figure stands upright on a staircase.
Full-size original

Google

Nano Banana 2: a woman in a white dress at the edge of a glass box holding a city, with upside-down buildings, waterfalls and two figures hanging upside down — Nano Banana 22752×1536 · 55 credits
The city fills a glass box above the sea. Two figures hang upside down under an island inside the box.
Full-size original

Nano Banana 2 Lite: a woman in a white dress walks into a diamond crystal holding a city, with a refracted second moon and a figure upside down under an island — Nano Banana 2 Lite1376×768 · 11 credits
Rain falls outside and beads on the glass. The crystal shows a refracted second moon, and a figure walks upside down under an island.
Full-size original

OpenAI

GPT Image 2: a woman in a white dress stands inside a curved crystal wall, with a city of upside-down buildings, a giant red moon and glowing whales — GPT Image 22048×1152 · 42 credits
The camera stands inside the crystal with her, at its wall, and the city surrounds them. There is no upside-down figure. This image is the start frame of our video comparison.
Full-size original

xAI

Grok Imagine: a woman in a white dress on a black platform inside a liquid glass form, with an upside-down city above and a glowing whale below — Grok Imagine1280×720 · 11 credits
An upside-down city hangs inside a liquid glass form, and a huge glowing whale swims under the platform. There are no staircases and no waterfalls.
Full-size original

Grok Imagine 2: a woman in a white dress beside a glass bubble with a city, rainbow light, a ghost copy of her and a figure hanging upside down — Grok Imagine 21792×1008 · 11 credits
Rainbow light plays on the bubble, and a ghost copy of the woman appears on its edge. A figure hangs upside down under a floating platform.
Full-size original

How we read the images

We checked every image, with full-size crops wherever a detail was small. A detail counts as drawn when it is clearly there. It counts as partly drawn when it is there but changed, for example a crystal that sits on the sea instead of floating above it. It counts as missing when we could not find it. For the waterfalls, we judged the direction from the shape of the water and its spray.

The prices are today's prices for the same settings, and each model's page lists all of its prices. We also ran two models that are no longer in Nodaro's model catalog, so this page does not show them.

This is one image per model, from one prompt. Another image from the same prompt can differ. Read the checklist as what each model drew this time, not as a ranking of the models.

Run the same test in Nodaro

Connect one Text node with your prompt to several Generate Image nodes. Choose a different model in each Generate Image node, and run the workflow. Each node shows its price on its Run button before it runs. To animate the image you like best, connect it to Generate Video nodes, as we did in the video comparison.

15 AI image models, one prompt: what each model drew