--- title: "Best AI Models: Personalization at scale" description: "I want to personalize creatives at scale: weighted model recommendations for personalization at scale." url: "https://img.ly/ai-benchmarks/use-cases/personalization-at-scale/" type: "benchmark-use-case" useCaseIds: ["U6"] --- > This is the markdown version of [Best AI Models: Personalization at scale](https://img.ly/ai-benchmarks/use-cases/personalization-at-scale/). 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). --- # Best AI models: Personalization at scale "I want to personalize creatives at scale" Thousands of variants per campaign, generated by automation rather than by hand: per-segment backgrounds, per-market scenes, per-recipient artwork. What matters: obedience to structured, machine-generated prompts, and unit cost that survives multiplication. ## Recommended models 1. [Nano Banana 2 Lite](https://img.ly/ai-benchmarks/models/nano-banana-2-lite.md): 4.26 / 5 weighted score (100% of the weight matrix covered, $0.020 per image) — best fit 2. [FLUX.2](https://img.ly/ai-benchmarks/models/flux-2.md): 4.10 / 5 weighted score (100% of the weight matrix covered, $0.013 per image) 3. [FLUX.2 [dev] Turbo](https://img.ly/ai-benchmarks/models/flux-2-turbo.md): 4.05 / 5 weighted score (100% of the weight matrix covered, $0.015 per image) 4. [Gemini 2.5 Flash Image](https://img.ly/ai-benchmarks/models/gemini-25-flash-image.md): 3.97 / 5 weighted score (100% of the weight matrix covered, $0.039 per image) 5. [Seedream 4.5](https://img.ly/ai-benchmarks/models/seedream-4-5.md): 3.91 / 5 weighted score (100% of the weight matrix covered, $0.048 per image) 6. [Seedream 5.0 Lite](https://img.ly/ai-benchmarks/models/seedream-5-lite.md): 3.88 / 5 weighted score (100% of the weight matrix covered, $0.020 per image) 7. [FLUX.2 [pro]](https://img.ly/ai-benchmarks/models/flux-2-pro.md): 3.86 / 5 weighted score (100% of the weight matrix covered, $0.040 per image) 8. [Qwen-Image](https://img.ly/ai-benchmarks/models/qwen-image.md): 3.75 / 5 weighted score (100% of the weight matrix covered, $0.030 per image) 9. [Nano Banana 2](https://img.ly/ai-benchmarks/models/nano-banana-2.md): 3.73 / 5 weighted score (100% of the weight matrix covered, $0.100 per image) 10. [GPT Image 1.5](https://img.ly/ai-benchmarks/models/gpt-image-1-5.md): 3.58 / 5 weighted score (100% of the weight matrix covered, $0.120 per image) 11. [Luma Photon](https://img.ly/ai-benchmarks/models/luma-photon.md): 3.53 / 5 weighted score (100% of the weight matrix covered, $0.019 per image) 12. [Recraft V3](https://img.ly/ai-benchmarks/models/recraft-v3.md): 3.38 / 5 weighted score (100% of the weight matrix covered, $0.040 per image) 13. [Ideogram 3.0](https://img.ly/ai-benchmarks/models/ideogram-v3.md): 3.34 / 5 weighted score (100% of the weight matrix covered, $0.060 per image) 14. [Nano Banana Pro](https://img.ly/ai-benchmarks/models/nano-banana-pro.md): 2.98 / 5 weighted score (100% of the weight matrix covered, $0.360 per image) 15. [Stable Diffusion 1.5](https://img.ly/ai-benchmarks/models/stable-diffusion-1-5.md): 2.98 / 5 weighted score (100% of the weight matrix covered, $0.010 per image) Nano Banana 2 Lite takes the top spot where this job puts its weight: prompt adherence at 4.7 / 5 (30% of the matrix), cost at 4.0 / 5 (25% of the matrix) and latency at 3.9 / 5 (15% of the matrix). ## Weights (pilot-0, provisional) Automation has no human retry loop, so prompt adherence leads: every variant that ignores its data-driven brief is a silent defect in someone's mailbox. Cost and latency follow because the same job runs thousands of times per campaign; per-image quality criteria stay light because variants share one reviewed template and inherit its typography and colors. - Prompt Adherence: 30% - Cost: 25% - Latency: 15% - Text Accuracy: 10% - Composition: 10% - Color Accuracy: 5% - Resolution: 5% Criteria whose scoring tier has not run yet are excluded and the remaining weights renormalized; the covered fraction is reported per model above. ## Relevant benchmarks - [Banner with Negative Space](https://img.ly/ai-benchmarks/prompts/n01-banner.md): Composition & Negative Space - [Vertical Story Background](https://img.ly/ai-benchmarks/prompts/n02-story-bg.md): Composition & Negative Space - [Centered Product Margin](https://img.ly/ai-benchmarks/prompts/n03-margin-product.md): Composition & Negative Space - [Structured JSON Scene](https://img.ly/ai-benchmarks/prompts/j01-json-scene.md): Structured Spec Adherence ## The last mile: what no model delivers Our structured-spec benchmark shows the strongest models follow machine-generated prompts remarkably well, and the budget tier makes per-variant cost viable. What generation cannot do is guarantee the deterministic parts of a variant: the recipient's name spelled correctly, the legal line, the logo placement, the brand hex. Automation pipelines that ship put those elements on the canvas as template layers and let the model supply only the interchangeable visual underneath. Generation routes through the [IMG.LY AI Gateway](https://img.ly/ai-gateway.md) and becomes a production-ready file on the [CE.SDK](https://img.ly/products/creative-sdk.md) canvas. --- ## 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.