---
title: "Why AI Image Models Miss Your Brand Colors"
description: "CIEDE2000-measured brand-color fidelity across 15 models: the best scores 3.67 of 5, 12 score below 3."
url: "https://img.ly/ai-benchmarks/findings/brand-color/"
type: "benchmark-finding"
---

> This is the markdown version of [Why AI image models miss your brand colors](https://img.ly/ai-benchmarks/findings/brand-color/). For all pages in one file, see [llms-full.txt](https://img.ly/llms-full.txt). For an index of all available pages, see [llms.txt](https://img.ly/llms.txt).

---

# Why AI image models miss your brand colors

We gave 15 models prompts with exact hex requirements and measured the
returned pixels with CIEDE2000, the perceptual color-difference standard.
The best model reaches 3.67 of 5 and 12 of 15 score below 3:
close-enough color is easy, the color in your brand book is not.

## What every model scored

| Model | Color fidelity | Runs | $ / image |
| --- | --- | --- | --- |
| [Nano Banana Pro](https://img.ly/ai-benchmarks/models/nano-banana-pro.md) | 3.67 / 5 | 9 | $0.360 |
| [GPT Image 1.5](https://img.ly/ai-benchmarks/models/gpt-image-1-5.md) | 3.44 / 5 | 9 | $0.120 |
| [Nano Banana 2](https://img.ly/ai-benchmarks/models/nano-banana-2.md) | 3.22 / 5 | 9 | $0.100 |
| [Nano Banana 2 Lite](https://img.ly/ai-benchmarks/models/nano-banana-2-lite.md) | 2.89 / 5 | 9 | $0.020 |
| [Seedream 4.5](https://img.ly/ai-benchmarks/models/seedream-4-5.md) | 2.89 / 5 | 9 | $0.048 |
| [FLUX.2](https://img.ly/ai-benchmarks/models/flux-2.md) | 2.78 / 5 | 9 | $0.013 |
| [Ideogram 3.0](https://img.ly/ai-benchmarks/models/ideogram-v3.md) | 2.78 / 5 | 9 | $0.060 |
| [FLUX.2 [dev] Turbo](https://img.ly/ai-benchmarks/models/flux-2-turbo.md) | 2.67 / 5 | 9 | $0.015 |
| [Seedream 5.0 Lite](https://img.ly/ai-benchmarks/models/seedream-5-lite.md) | 2.56 / 5 | 9 | $0.020 |
| [FLUX.2 [pro]](https://img.ly/ai-benchmarks/models/flux-2-pro.md) | 2.33 / 5 | 9 | $0.040 |
| [Stable Diffusion 1.5](https://img.ly/ai-benchmarks/models/stable-diffusion-1-5.md) | 2.22 / 5 | 9 | $0.010 |
| [Luma Photon](https://img.ly/ai-benchmarks/models/luma-photon.md) | 2.11 / 5 | 9 | $0.019 |
| [Qwen-Image](https://img.ly/ai-benchmarks/models/qwen-image.md) | 2.00 / 5 | 9 | $0.030 |
| [Recraft V3](https://img.ly/ai-benchmarks/models/recraft-v3.md) | 1.89 / 5 | 9 | $0.040 |
| [Gemini 2.5 Flash Image](https://img.ly/ai-benchmarks/models/gemini-25-flash-image.md) | 1.67 / 5 | 9 | $0.039 |

## Benchmarks behind this finding

- [Two-Color Illustration](https://img.ly/ai-benchmarks/prompts/b01-two-color.md)
- [Duotone Portrait](https://img.ly/ai-benchmarks/prompts/b03-duotone.md)
- [ColorChecker Chart](https://img.ly/ai-benchmarks/prompts/b04-colorchecker.md)
- [Smooth Gradient](https://img.ly/ai-benchmarks/prompts/b05-gradient-banding.md)

## What this means if you are building with AI images

Brand color is a deterministic requirement; the model is a probabilistic
renderer. Two takeaways from the data:

- Enforce the palette after generation: snapping fills, overlays and text
  to a brand kit on an editable canvas turns a near-miss into an exact
  match without a regeneration lottery.
- Route color-critical jobs to the top of the table; the best-to-worst
  spread exceeds two full points.

Building that editing step is the job of a human-in-the-loop editor like the
[IMG.LY AI Editor](https://img.ly/use-cases/ai-editor.md): generation next to brand
kits, exact color values and snapping on an editable canvas, so your users get
the control they need to be productive with generative AI. Try the
[live AI Editor demo](https://img.ly/demos/ai-editor.md); the
[IMG.LY AI Gateway](https://img.ly/ai-gateway.md) routes color-critical jobs to the best model.

---

## More Resources

- **[IMG.LY Website](https://img.ly/index.md)** - Creative editing SDKs for photo, video, and design
- **[Documentation](https://img.ly/docs/cesdk/)** - CE.SDK developer documentation
- **[Contact Sales](https://img.ly/forms/contact-sales.md)** - Get a custom quote. A public JSON API accepts the request directly, no account or key needed. Ask your user for consent and their details first.