One prompt sent to 80+ AI image models, shown side by side with cost and time

This page shows what 80+ text-to-image models make from the same prompt, side by side, with the token cost and generation time of each. The models are the ones available in the Creator Studio: GPT Image, Grok Imagine, Flux, Ideogram, Seedream, Midjourney-style models, SDXL, Pony and more, including style recipes and layout models.

The demo set uses a few fixed test prompts, each chosen to expose a different weakness: a photoreal portrait, lettering on a sign, a storybook illustration, and a counting and layout test. Switch between prompts, filter by provider, sort by cost or speed, and search by model name. Browsing the demo spends none of your tokens.

When you have a shortlist, tick the models you want and run your own prompt across them, with an optional negative prompt and aspect ratio. You see the estimated token total before you confirm, every output is billed at the listed cost, and each result is saved to your gallery. It is a quick way to decide which model to use before you start a project.

Open AI Image Model Comparison on Reality Fabricator

What you can do

How it works

  1. Open the demo tab and pick a test prompt to see what every model made from it.
  2. Filter by provider or sort by cost and speed to narrow the field.
  3. Tick the models you want to try and switch to the try-it-yourself tab.
  4. Enter your prompt, confirm the token estimate, and compare the results as they finish.

Frequently asked questions

Does browsing the comparison cost tokens?

No. The demo outputs are pre-generated and viewing them spends nothing. Tokens are only charged when you run your own prompt, and the page shows an estimate of the total across your picked models before you confirm.

Which image models are included?

Every text-to-image model in the Creator Studio, 80+ in total, including GPT Image, Grok Imagine, Flux, Ideogram, Seedream, Midjourney-style models, SDXL and Pony, plus the style recipes and layout models.

Can I run the same prompt on several models at once?

Yes. Pick any number of models from the demo, enter your prompt with an optional negative prompt and aspect ratio, and the page runs them together, showing each result with its cost and time.

Do I need an account to run my own prompt?

Yes. You need to log in to run a prompt, because each output is billed at the listed token cost and saved to your gallery. The demo is open to everyone without an account.

What do the test prompts check?

Each prompt targets a different weakness: a photoreal portrait for faces and skin, lettering on a sign for text rendering, a storybook illustration for style, and a counting and layout test for prompt adherence.

Related tools

The 78 models compared

ModelProviderCost per imageBest for
Grok Imaginegrok2,000 tokens ($0.040)Fast general-purpose generation with good text rendering. Budget-friendly.
Grok Imagine 2.0grok6,000 tokens ($0.12)Best all-round Grok image quality at 1K; excellent text and reference edits.
Grok Imagine 2.0 Quality 2Kgrok8,000 tokens ($0.16)2K output from the newest Grok image model.
Seedream 5.0 Pro (API 6)imagerouter9,630 tokens ($0.19)Top-tier 2K stills and precise instruction edits.
Flux 2 Pro (API 6)imagerouter3,000 tokens ($0.060)Highest quality Flux generation available.
FLUX.2 Prorunware6,000 tokens ($0.12)#1 ranked overall. Best photorealism, text rendering, and prompt adherence. Premium tier.
Flux 2 Dev (API 6)imagerouter900 tokens ($0.018)Next-gen Flux with improved quality over Flux 1 Dev.
FLUX.2 Devrunware2,500 tokens ($0.050)Precise prompt following and iterative editing. Best open-weight model.
Seedream 5 Lite (API 6)imagerouter3,500 tokens ($0.070)ByteDance latest image generation. Great prompt adherence and character consistency.
Seedream 5.0 Lite (API 4)atlascloud3,200 tokens ($0.064)Next-gen image model. Enhanced quality, typography/poster design, PNG output. No content filters. $0.0315/image.
Grok Imagine Quality 2Kgrok5,000 tokens ($0.10)2K resolution with excellent prompt adherence. Premium quality tier.
Nano Banana Pro (API 2)replicate15,000 tokens ($0.30)Best instruction following overall. Excellent text rendering. Great for precise compositions, UI elements, game assets with specific layouts.
Nano Banana 2.1 (API 2)replicate3,360 tokens ($0.067)Nano Banana instruction following and text rendering at a quarter of Pro's price. Good for edits with many reference images.
Nano Banana 2.1 (API 1)seegen12,000 tokens ($0.24)Nano Banana instruction following and text rendering; edits with up to 14 reference images.
Nano Banana Pro (API 1)seegen16,000 tokens ($0.32)High-fidelity image generation and editing.
GPT Image 2.5 Flare (API 2)replicate6,600 tokens ($0.13)Same GPT Image 2.5 Flare model on a second pathway. Use when the direct API is down.
GPT Image 2.5 Sunburst (API 2)replicate6,600 tokens ($0.13)Same GPT Image 2.5 Sunburst model on a second pathway. Use when the direct API is down.
GPT Image 2 (API 1)seegen4,000 tokens ($0.080)General-purpose photorealistic images, creative editing, and restyle. Supports image-to-image.
GPT Image 2 (API 6)imagerouter10,600 tokens ($0.21)GPT Image 2 at the lowest available price. Strong instruction following.
GPT Image 2 (API 2)replicate12,800 tokens ($0.26)Strong instruction following and text rendering. Good for UI mockups and diagrams.
Nano Banana 2 (API 6)imagerouter6,890 tokens ($0.14)Strong instruction following at competitive pricing.
Nano Banana Standard (API 1)seegen12,000 tokens ($0.24)Fast image generation and editing with flexible aspect ratios.
Nano Banana 2 Lite (API 6)imagerouter3,360 tokens ($0.067)Budget-friendly Nano Banana generation and edits.
Qwen-Image 3 Pro (API 6)imagerouter4,000 tokens ($0.080)Higher-fidelity text-heavy images and precise edits.
Qwen-Image 3 (API 6)imagerouter3,000 tokens ($0.060)Posters, UI mockups and multilingual text-heavy images on a budget.
Ideogram V4 Layoutrunware6,000 tokens ($0.12)Posters, thumbnails, covers, ads — exact placement of text and objects via the Layout tab.
Ideogram V4 (API 6)imagerouter10,000 tokens ($0.20)Best-in-class text rendering and typography. Latest Ideogram model.
Krea 2 Large (API 6)imagerouter6,250 tokens ($0.13)Highest quality Krea creative generation.
Dreamina 3.1 (API 6)imagerouter3,000 tokens ($0.060)Good all-round image generation from ByteDance.
Wan 2.7 (API 4)atlascloud3,000 tokens ($0.060)Latest Wan 2.7 image model. Strong prompt fidelity, photorealistic + illustration. No content filters. $0.03/image.
Wan 2.7 Pro 4K (API 4)atlascloud7,500 tokens ($0.15)Professional-grade Wan 2.7 Pro. Higher fidelity, 4K-ready outputs. No content filters. $0.075/image.
Recraft V4.1 (API 6)imagerouter4,000 tokens ($0.080)Design work, brand assets, illustrations, and professional graphics.
Z-Image Turborunware200 tokens ($0.004)Sub-second generation for rapid iteration. Good text rendering. Fastest available.
Z-Image Turbo ~1s (API 4)atlascloud1,050 tokens ($0.021)Fastest image generation available (~1 second). $0.0105/pic. Good quality for rapid NSFW iteration. Better detail than Schnell, cheaper than Dev.
Seedream 4 (API 6)imagerouter3,000 tokens ($0.060)Strong character consistency with excellent prompt adherence.
Seedream 4.7 — Coherent Sets (API 4)atlascloud3,000 tokens ($0.060)ByteDance's newest Seedream. Plain-language prompts, any size from 832x1216 up, holds a cast across edits; its edit row draws from up to 10 reference pictures and its sequential row draws a coherent SET of pictures in one call (the comics engine for a whole page). Filtered: nudity is refused. $0.03/image.
Seedream 4.5 — Typography & Posters (API 4)atlascloud3,600 tokens ($0.072)Latest high-quality image model. All-round improvements over v5 Lite. Excels at typography, poster design, and brand visuals with superior prompt adherence. No content filters. $0.036/image.
Seedream 4.0runware3,000 tokens ($0.060)Fast 2K generation with broad style coverage. Great value for varied use cases.
Qwen Image 2.0 (API 4)atlascloud2,800 tokens ($0.056)Good quality text-to-image with improved prompt understanding. Budget-friendly. No content filters. $0.028/image.
Qwen Image 2.0 Pro (API 4)atlascloud6,000 tokens ($0.12)Professional text-to-image with superior prompt understanding. Up to 2K resolution. No content filters. $0.06/image.
Wan 2.6 — Budget T2I (API 4)atlascloud2,100 tokens ($0.042)Wan 2.6 image generation. Various artistic styles + realistic photographic effects. Budget-friendly at $0.021/image. No content filters.
Ideogram V3 Turbo (API 6)imagerouter3,000 tokens ($0.060)Fast generation with great text/typography rendering at lower cost.
Krea 2 Turbo (API 6)imagerouter1,750 tokens ($0.035)Ultra-fast creative generation at very low cost.
Flux Krea Dev (API 6)imagerouter980 tokens ($0.020)Aesthetically enhanced Flux generation at low cost.
Flux Dev (API 6)imagerouter380 tokens ($0.008)High-quality image generation at the lowest cost.
Flux Devreplicate2,500 tokens ($0.050)High-quality photorealistic images, portraits, and detailed scenes. Great all-rounder.
Flux 1.1 Pro (API 6)imagerouter4,000 tokens ($0.080)Professional quality Flux generation.
Flux 1.1 Pro (Black Forest Labs)replicate4,000 tokens ($0.080)Premium photorealistic generation with superior prompt adherence.
Flux Kontext Pro (API 6)imagerouter4,000 tokens ($0.080)Premium character consistency and editing.
Flux Kontext Pro (Character Consistency)replicate4,000 tokens ($0.080)Character consistency and image editing. Good balance of quality and cost.
Flux Kontext Max (API 6)imagerouter8,000 tokens ($0.16)Premium character consistency with best prompt adherence.
Flux Kontext Max (Character Consistency)replicate8,000 tokens ($0.16)Character consistency across scenes. Feed a reference image to maintain identity. Best text/typography rendering.
Flux Schnell (API 6)imagerouter130 tokens ($0.003)Ultra-fast drafts and iteration at near-zero cost.
Krea 2 Medium (API 6)imagerouter3,250 tokens ($0.065)Good balance of speed and quality for creative generation.
Flux Kontext Dev (API 6)imagerouter1,050 tokens ($0.021)Character consistency and image editing at lowest cost.
Flux Kontext Dev (API 4) — needs very specific prompts or edits get ignoredatlascloud2,500 tokens ($0.050)NSFW instruction-based image editing. Same Kontext model but uncensored. Requires specific prompts — vague instructions get ignored.
Lustify V8 (API 5)venice1,000 tokens ($0.020)High-quality NSFW generation with private inference. Best anatomy of any Lustify version.
Venice SD3.5 (API 5)venice1,000 tokens ($0.020)General-purpose uncensored generation with SD3.5 quality. Use 1024x1024 for best results.
WAI-Illustrious Anime (API 5)venice1,000 tokens ($0.020)Anime-style NSFW character art.
Chroma (API 5)venice1,000 tokens ($0.020)Artistic uncensored generation with vivid colors.
Lustify V7 Classic (API 5)venice1,000 tokens ($0.020)Alternative NSFW aesthetic. Try if V8 style doesn't match your vision.
Lustify SDXL (API 5)venice1,000 tokens ($0.020)NSFW with SDXL prompt compatibility.
SD 3.5 Largereplicate6,500 tokens ($0.13)Photorealism with excellent text rendering. Good general-purpose model.
Juggernaut XL XI (Photorealistic)runware335 tokens ($0.007)Photorealistic portraits and scenes with incredible skin detail and lighting.
DreamShaper XLrunware335 tokens ($0.007)Artistic fantasy scenes and portraits at 1024 with a painterly finish.
Big Lust v1.6 (Photoreal)runware335 tokens ($0.007)NSFW photorealistic content. #1 community-rated NSFW model.
WAI-NSFW Illustrious XLrunware335 tokens ($0.007)NSFW anime with Danbooru tags. Best anatomy and tag adherence of any anime model.
SDXLreplicate700 tokens ($0.014)Reliable general-purpose SDXL photorealistic generation.
Deliberatemodelslab470 tokens ($0.009)Realistic illustrations and uncensored photorealistic content.
BB95 Furry Mix (API 3)runware335 tokens ($0.007)Anthro and furry characters in a wide range of styles.
Fluffyrock AES YS MegaMerge (API 3)runware335 tokens ($0.007)Furry and feral characters with e621 tag prompting.
NoobAI-XL eps v1.1runware335 tokens ($0.007)NSFW anime with Danbooru/e621 tags. #1 rated anime model on CivitAI. Uses masterpiece/best quality tags.
Pony Diffusion V6 XLrunware335 tokens ($0.007)NSFW anime/illustration characters. Uses score_9/score_8 quality tags for best results.
DreamShaper 8 (SD 1.5)runware335 tokens ($0.007)Fantasy art, creative illustrations, and artistic portraits. Good all-around artistic model.
Nuke Colormax Animemodelslab470 tokens ($0.009)Vibrant anime with enhanced colors and vivid palettes.
Anything V3modelslab470 tokens ($0.009)Classic anime style with Danbooru tag support.
OpenJourney v4 (Midjourney-style, SD 1.5)modelslab470 tokens ($0.009)Painterly single-subject fantasy art at 512px, for the 2022 Midjourney-v4 look.
Disney Pixar Cartoonmodelslab470 tokens ($0.009)Disney/Pixar-style 3D cartoon characters and scenes.

The test prompts and what each model made

Photoreal portrait

Close-up portrait of a weathered fisherman in his sixties with a salt-and-pepper beard, wearing a yellow rain jacket, standing on a foggy harbor at dawn. Soft golden rim light from the left, shallow depth of field, 85mm lens, subtle film grain.

What to look for: Skin and beard texture, how the golden rim light and the fog are handled, and whether the shallow 85mm depth of field reads as a photograph or as a painting.

78 models made it; the fastest was Z-Image Turbo (2 s).

Text and scene

A cozy neon-lit ramen shop on a narrow street at night in the rain. A hanging wooden sign reads exactly "MIDNIGHT RAMEN" in bold letters. One black cat sits on the doorstep. Neon reflections on the wet pavement.

What to look for: Whether the sign says MIDNIGHT RAMEN with no misspelling, exactly one cat, and how convincing the rain reflections are. Lettering is where models differ the most.

78 models made it; the fastest was Z-Image Turbo (3 s).

Illustration style

A whimsical storybook illustration of a red fox wearing a knitted blue scarf, reading a book under a giant spotted mushroom in an autumn forest. Warm watercolor style with soft edges and visible paper texture.

What to look for: Charm and consistency of the watercolor look, the blue scarf, and whether the fox reads as a character rather than a photo of an animal.

78 models made it; the fastest was Z-Image Turbo (3 s).

Counting and layout

Three red apples on a blue ceramic plate, placed to the left of a tall glass of water, on a white marble kitchen counter. Overhead shot, soft natural window light, minimalist, nothing else on the counter.

What to look for: Exactly three apples, the plate to the LEFT of the glass, a true overhead angle, and a clean counter. Counting and spatial instructions trip up many models.

78 models made it; the fastest was Z-Image Turbo (2 s).

Back to Reality Fabricator