Pick a model, run it in bulk

BulkImagen runs 16 image models from 6 labs on one credit balance — 1 to 20 credits per image, and no provider API keys of your own. Compare what each one costs and supports, then send a whole prompt list through it.

Not sure which one? Start here

By job

Pick by what the batch is for

Start from the job, not the model name. Each card lists the models that do that job best here, with the price you would pay per image.

Text, posters and labels

Headlines, packaging and signs that have to be spelled right, in Chinese as well as English.

Photoreal scenes and products

Skin, light and materials that pass for a photograph, for product shots and lifestyle scenes.

Same character, same product

Attach a stack of reference images so every row of the batch keeps the same face, product or style.

Cut-outs on a transparent PNG

Stickers, icons and product cut-outs that come back with a real alpha channel, ready to drop on any background.

Big batches on a budget

The lowest price per image, for when you want hundreds of drafts to choose from.

Print and large formats

The least expensive ways to get a 4K image, for posters, print and anything that gets cropped.

Side by side

Every model, every spec

These numbers are read from the generation registry, so they are what a batch actually charges. Credits are billed per finished image, at the rate for the resolution you pick.

ModelMade byCredits / imageResolutionsAspect ratiosReference imagesPrompt limit
OpenAIGPT Image 2OpenAI1K 5 · 2K 10 · 4K 201K · 2K · 4K15up to 420,000 chars
OpenAIGPT Image 2.5OpenAI1K 5 · 2K 10 · 4K 161K · 2K · 4K12up to 820,000 chars
GeminiNano Banana ProGoogle1K 5 · 2K 6 · 4K 101K · 2K · 4K10up to 410,000 chars
GeminiNano Banana 2Google1K 4 · 2K 6 · 4K 101K · 2K · 4K14up to 420,000 chars
GeminiNano Banana 2 LiteGoogle31K14up to 420,000 chars
GeminiNano Banana 2.1Google1K 4 · 2K 6 · 4K 101K · 2K · 4K14up to 420,000 chars
ByteDanceSeedream 4.5ByteDance2K 6 · 4K 62K · 4K8up to 43,000 chars
ByteDanceSeedream 5.0 LiteByteDance2K 5 · 4K 52K · 4K8up to 43,000 chars
ByteDanceSeedream 5.0 ProByteDance1K 6 · 2K 121K · 2K8up to 45,000 chars
ByteDanceSeedream 5.0 FlashByteDance1K 3 · 2K 31K · 2K8up to 105,000 chars
GrokGrok Imagine 2.0xAI41K5up to 15,000 chars
FluxFlux 2 ProBlack Forest Labs1K 5 · 2K 61K · 2K7up to 85,000 chars
QwenQwen Image 3.0 ProAlibaba1K 6 · 2K 101K · 2K8up to 35,000 chars
QwenQwen Image 3.0Alibaba1K 4 · 2K 41K · 2K8up to 35,000 chars
QwenQwen Image 2.1Alibaba1K 4 · 2K 61K · 2K9up to 105,000 chars
QwenZ-ImageAlibaba11K5Text only1,000 chars

Aspect-ratio counts leave out the automatic option some models offer. Reference-image limits apply per call. If an image ends up failing, its credits go back to your balance.

Specs narrow it down. Your prompt decides.

Run the same prompt on several of these models side by side and see which one actually fits your job before you batch.
Compare models on your prompt
By family

6 families, what sets each apart

Models from the same lab share a look. Here is what each family is known for, and how the models inside it differ.

OpenAI

OpenAI

Accurate text, transparent backgrounds
  • GPT Image 2PopularSharp, prompt-faithful renders with legible text1K52K104K204 referencesTransparent BG
  • GPT Image 2.5NewNewest generation, sharper text and detail1K52K104K168 referencesTransparent BG
Read the guide
Gemini

Google

Fast and versatile
Read the guide
ByteDance

Seedream

Photoreal detail
Read the guide
Qwen

Qwen

Chinese & English text, lowest prices
  • Qwen Image 3.0 ProSharpest Qwen, strong at text in images1K62K103 references
  • Qwen Image 3.0Crisp Chinese and English text, low price1K42K43 references
  • Qwen Image 2.1Reliable all-rounder, many reference images1K42K610 references
  • Z-ImageCheapest way to fill a big batch, text only1K1Text only
Read the guide
Grok

Grok

Stylized looks
Read the guide
Flux

Flux

Photoreal, many references
  • Flux 2 ProPhotoreal detail, takes a whole mood board1K52K68 references
Read the guide
From the blog

Model comparisons, tested

All comparisons

Model questions

Something we didn't cover? Email support@bulkimagen.com

  • For large, cost-sensitive runs, Z-Image at 1 credit an image is the cheapest option here. For everyday work the form defaults to GPT Image 2. If you need 4K, GPT Image 2, GPT Image 2.5, Nano Banana Pro, Nano Banana 2, Nano Banana 2.1, Seedream 4.5, and Seedream 5.0 Lite offer a 4K tier.
  • Z-Image, at 1 credit per image. Across the catalogue, rates run from 1 to 20 credits an image depending on the model and the resolution you choose.
  • GPT Image 2, GPT Image 2.5, Nano Banana Pro, Nano Banana 2, Nano Banana 2.1, Seedream 4.5, and Seedream 5.0 Lite. The rest top out at 1K or 2K — the table above lists the tiers each model offers, and the price of every tier.
  • 15 of the 16. Seedream 5.0 Flash takes the most at 10 per call. Z-Image is text-only, and a batch that attaches references to it is rejected before any credits are charged.
  • Per finished image, at the model's rate for the resolution you pick — 1 to 20 credits. There is no per-batch fee and no monthly minimum, and credits for images that fail are returned automatically.
  • Yes. The model is chosen per batch on the form, not tied to your plan, so a common pattern is a cheap first pass to find the prompts that work, then a rerun of the keepers on a stronger model.
  • No. BulkImagen holds the provider keys and bills you in credits, so every model in the table is available as soon as you sign in — including ones whose APIs are otherwise waitlisted or region-limited.