Key takeaways
- Recraft V4.1 Flash is the cheapest image model listed, at $0.007 an image; Recraft V4 Pro and the Pro vector styles cost the most, at $0.25 to $0.30.
- Seedream models cost $0.03 to $0.04 an image, and Grok Imagine costs $0.02 or $0.04 to $0.08 depending on version, size and quality.
- Every model is billed per image at the provider's list price plus your platform fee, with no subscription.
- All of them take the OpenAI images request, so client.images.generate works by changing the base URL and the model.
Image models are priced per image, but the price depends on the model, the size, the quality and sometimes the style. Below are the list prices for the Recraft, Seedream and Grok Imagine models on Inference Bureau, as verified against the upstream model list on Oct 19, 2026, with FLUX and Muse Image for comparison.
Prices are in US dollars per image, before your platform fee.
Recraft
Recraft models charge more for vector output. The second column is the price with style: "vector_illustration".
| Model | Per image | Vector illustration |
|---|---|---|
recraft/recraft-v4.1-flash | $0.007 | n/a |
recraft/recraft-v2 | $0.022 | $0.044 |
recraft/recraft-v4.1 | $0.035 | $0.08 |
recraft/recraft-v4.1-utility | $0.035 | $0.08 |
recraft/recraft-v3 | $0.04 | $0.08 |
recraft/recraft-v4 | $0.04 | $0.08 |
recraft/recraft-v4.1-pro | $0.21 | $0.30 |
recraft/recraft-v4.1-utility-pro | $0.21 | $0.30 |
recraft/recraft-v4-pro | $0.25 | $0.30 |
Seedream
ByteDance's Seedream models have one price per image at any size. None of them train on your prompts.
| Model | Per image |
|---|---|
bytedance/seedream-4.0 | $0.03 |
bytedance/seedream-5.0-lite | $0.035 |
bytedance/seedream-5.0-pro | $0.035, plus a small charge per prompt token |
bytedance/seedream-4.5 | $0.04 |
Grok Imagine
Both Grok Imagine models run with zero data retention and no training on your prompts. Grok Imagine 2.0 is priced by size and quality.
| Model | Size and quality | Per image |
|---|---|---|
spacexai/grok-imagine-image | any | $0.02 |
spacexai/grok-imagine-image-2.0 | default size, low quality | $0.04 |
spacexai/grok-imagine-image-2.0 | default size and quality | $0.06 |
spacexai/grok-imagine-image-2.0 | 2048×2048, low quality | $0.06 |
spacexai/grok-imagine-image-2.0 | 2048×2048 | $0.08 |
For comparison
| Model | Per image |
|---|---|
bfl/flux-3-image-klein | $0.008 to $0.036 by size ($0.009 at 1024×1024) |
meta/muse-image-1.0 | $0.01 |
bfl/flux-3-image | $0.041 to $0.607 by size ($0.048 at 1024×1024) |
Which one to pick
- Lowest cost per image: Recraft V4.1 Flash at $0.007, then FLUX 3 Klein and Muse Image at about a cent.
- Vector output: any Recraft model with
vector_illustration, from $0.044 on Recraft V2. - Prompts that must not be kept: Grok Imagine and Muse Image run with zero data retention.
- One flat price at any size: Seedream.
Price is only half of it. Try the same prompt on two or three models before you choose: at these prices, a test of a dozen images costs well under a dollar.
Calling them
Every image model takes the OpenAI images request at POST /v1/images/generations, so the OpenAI SDK works with a base URL change:
import OpenAI from "openai";
const client = new OpenAI({ baseURL: "https://api.inferencebureau.com/v1", apiKey: process.env.IB_API_KEY });
const res = await client.images.generate({
model: "recraft/recraft-v4.1-flash",
prompt: "A lighthouse at dusk",
size: "1024x1024",
});Images come back as base64 in data[].b64_json, since they aren't stored. You can ask for up to four images per request with n, and pass a provider's own options, such as Recraft's style, in providerOptions.
How images are billed
Each image is billed at the provider's list price plus your platform fee, which is 7.5% today. Before the call, your balance holds the most the request can cost: n images at the price of the size you asked for, or the most expensive size, quality or style your options allow. The request then settles at what the provider actually billed, and the receipt is in Logs like any other request.
Frequently asked questions
What is the cheapest image generation model?
Of the models on Inference Bureau, Recraft V4.1 Flash is the cheapest at $0.007 an image before the platform fee. FLUX 3 Klein starts at $0.008 for small images, and Muse Image costs $0.01.
Can I use the OpenAI SDK for Recraft, Seedream and Grok Imagine?
Yes. All of them take the OpenAI images request at /v1/images/generations, so client.images.generate works once you change the base URL, the API key and the model.
Are images stored?
No. Images are returned as base64 in the response and are not stored, so the url response format is refused.