---
title: "How Reliably Do AI Models Render Text?"
description: "Transcription-measured text accuracy across 15 models: 3 models rendered every required string exactly, while the legacy baseline garbles every word."
url: "https://img.ly/ai-benchmarks/findings/text-reliability/"
type: "benchmark-finding"
---

> This is the markdown version of [AI text rendering: close is not shippable](https://img.ly/ai-benchmarks/findings/text-reliability/). 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).

---

# AI text rendering: close is not shippable

We required exact strings in 4 typography benchmarks and transcribed
what 15 models actually rendered. 3 models, led by FLUX.2 [pro], rendered every required string in this suite exactly. One wrong character
makes a headline unusable, and baked-in text cannot be fixed without
regenerating the whole image.

## What every model scored

| Model | Text accuracy | Runs | $ / image |
| --- | --- | --- | --- |
| [FLUX.2 [pro]](https://img.ly/ai-benchmarks/models/flux-2-pro.md) | 5.00 / 5 | 24 | $0.040 |
| [GPT Image 1.5](https://img.ly/ai-benchmarks/models/gpt-image-1-5.md) | 5.00 / 5 | 24 | $0.120 |
| [Nano Banana Pro](https://img.ly/ai-benchmarks/models/nano-banana-pro.md) | 5.00 / 5 | 24 | $0.360 |
| [Gemini 2.5 Flash Image](https://img.ly/ai-benchmarks/models/gemini-25-flash-image.md) | 4.96 / 5 | 24 | $0.039 |
| [Nano Banana 2](https://img.ly/ai-benchmarks/models/nano-banana-2.md) | 4.96 / 5 | 24 | $0.100 |
| [FLUX.2 [dev] Turbo](https://img.ly/ai-benchmarks/models/flux-2-turbo.md) | 4.92 / 5 | 24 | $0.015 |
| [FLUX.2](https://img.ly/ai-benchmarks/models/flux-2.md) | 4.92 / 5 | 24 | $0.013 |
| [Nano Banana 2 Lite](https://img.ly/ai-benchmarks/models/nano-banana-2-lite.md) | 4.92 / 5 | 24 | $0.020 |
| [Seedream 4.5](https://img.ly/ai-benchmarks/models/seedream-4-5.md) | 4.88 / 5 | 24 | $0.048 |
| [Seedream 5.0 Lite](https://img.ly/ai-benchmarks/models/seedream-5-lite.md) | 4.88 / 5 | 24 | $0.020 |
| [Qwen-Image](https://img.ly/ai-benchmarks/models/qwen-image.md) | 4.83 / 5 | 24 | $0.030 |
| [Ideogram 3.0](https://img.ly/ai-benchmarks/models/ideogram-v3.md) | 4.71 / 5 | 24 | $0.060 |
| [Luma Photon](https://img.ly/ai-benchmarks/models/luma-photon.md) | 4.13 / 5 | 24 | $0.019 |
| [Recraft V3](https://img.ly/ai-benchmarks/models/recraft-v3.md) | 4.00 / 5 | 24 | $0.040 |
| [Stable Diffusion 1.5](https://img.ly/ai-benchmarks/models/stable-diffusion-1-5.md) | 1.17 / 5 | 24 | $0.010 |

Reputation check: Ideogram, the model known for text rendering, lands at #12 of 15.
The general flagships have absorbed the lead that made text specialists special.

## Benchmarks behind this finding

- [Wordmark](https://img.ly/ai-benchmarks/prompts/t04-wordmark.md)
- [Poster Headline](https://img.ly/ai-benchmarks/prompts/t01-poster-headline.md)
- [Product Label](https://img.ly/ai-benchmarks/prompts/t02-dense-label.md)
- [CJK Neon Sign](https://img.ly/ai-benchmarks/prompts/t05-cjk-sign.md)

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

- Keep headlines as editable text layers: generate the visual, set the
  words as real typography on a canvas. A wrong word becomes a two-second
  fix, fonts stay licensed and brand-exact, and localization is a string
  swap instead of a re-prompt per language.
- When text must be baked into the image, route to the top of the table;
  below the top tier misspelling is routine.

Closing that gap is the job of a human-in-the-loop editor like the
[IMG.LY AI Editor](https://img.ly/use-cases/ai-editor.md): generated visuals next to
real, editable typography, so a wrong word is a keystroke fix and 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 text-critical jobs to the top of the table.

---

## 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.