Stability AI (Stable Diffusion) API Pricing: Full Breakdown of Costs (Jun 2026)

This guide breaks down what every Stability AI API model costs, what drives your bill up, how to bring it down, and how to run Stability models for free (see also our free Stability AI API tutorial).

How much does the Stability AI API cost?

Stability AI's flagship generation service, Stable Image Ultra, costs $0.08 per image. The cheapest generation service, Stable Image Core, costs $0.03 per image. Stability bills in credits rather than tokens: 1 credit = $0.01, and each API call deducts a fixed number of credits based on the service you use. There is no separate input or output charge, because you are paying per generated asset, not per word.

Prices are quoted in credits. Generation services use the flat per-image rates above, while editing, upscaling, control, and audio services are billed separately at their own credit rates (covered below). Enterprise and bulk pricing is available through Stability's sales team at volume.

How Stability AI API pricing works

Stability bills per operation in credits. Each API call deducts a fixed number of credits for the service you use, with no separate input or output charge.

Credits and what one credit buys

One credit equals $0.01. Credits are sold in advance and deducted as you call the API. A 1,000-credit top-up costs $10. You can check your balance and buy more on your account page. Creating your account with Google's social login grants 25 free credits.

Flat per-generation pricing on current models

The current Stable Image services (Core, Ultra) and the Stable Diffusion 3.5 family charge a fixed credit cost per successful generation, regardless of resolution within the supported range or the number of diffusion steps. A Core image is 3 credits whether you generate it at one aspect ratio or another. This is simpler to forecast than the legacy SDXL and SD1.6 endpoints, which varied cost by resolution and step count.

Editing, upscaling, and control are billed as separate operations

If you generate an image and then edit or upscale it, you pay for each step. A generation, an inpaint, and an upscale are three separate credit charges on the same asset. The editing and control services (Erase, Inpaint, Remove Background, Search and Replace, Sketch, Structure, Style) are 5 credits each. Upscaling ranges from 2 credits (Fast) to 40 (Conservative) to 60 (Creative).

Audio and 3D are billed per generation

Text-to-Audio and Audio-to-Audio are 20 credits per generation. Fast 3D is 10 credits. These are flat per-call rates, the same model as image generation.

What makes your bill higher than expected

Creative Upscale is 60 credits

Creative Upscale costs 60 credits, or $0.60 per image. That is 20 times the cost of a Core generation. If you run it on every output rather than on selected finals, upscaling can become the largest line on your bill. Fast Upscale is 2 credits and Conservative is 40, so the choice of upscale mode matters more than the choice of generation model.

Multi-step workflows stack

A production pipeline often generates an image, removes the background, and upscales it. At Core rates that is 3 + 5 + 40 = 48 credits, or $0.48, for one finished asset, against $0.03 for the raw generation alone. Budget for the whole workflow, not the generation step.

There is no recurring free allowance

The 25 Google-signup credits are one time. There is no monthly free quota on the API. Once they are spent, every call costs credits, so there is no free tier to fall back on for low-volume production use.

How to reduce Stability AI API costs

1. Match the model to the job

Model choice is the largest lever. Core at $0.03 costs less than half of Ultra at $0.08. For thumbnails, social posts, drafts, and high-volume work, Core or SD3.5 Medium is usually enough. Reserve Ultra for final assets where detail and text rendering matter. Prototype on Turbo or Core, then regenerate only the keepers on Ultra.

2. Self-host the open weights if you are under $1M in revenue

Stability releases SD3.5 and SDXL as open-weight models. Under the Stability AI Community License, organizations generating under $1M in annual revenue can run them commercially at no license cost. Self-hosting removes the per-image credit charge entirely; you pay only for GPU compute. At high volume this is the cheapest option, and it keeps image data on your own infrastructure. Above $1M in revenue you need an Enterprise License.

3. Use the right upscale mode, or skip it

Creative Upscale (60 credits) produces the most transformation but costs 20 times a Core generation. Conservative (40) and Fast (2) cover most needs. Generate at the resolution you need where possible, and upscale only the outputs you keep.

4. Reuse generated assets instead of regenerating

If your application produces similar images repeatedly, store and reuse them rather than calling the API for each request. Deduplicating common prompts removes redundant credit spend on identical or near-identical outputs.

5. Negotiate bulk pricing above roughly 10,000 images per month

Stability offers volume discounts and custom pricing through its sales team. Teams generating thousands of images per month should ask, since the per-image rate can drop below list at committed volume.

Can you use the Stability AI API for free?

Puter.js: the User-Pays model

Puter.js is a JavaScript library that lets you add Stability AI models to your app with no API key, no backend, and no bill to you as the developer. It works on the User-Pays model: each user of your app covers their own image generation through their Puter account, so your costs stay at zero no matter how many users you have.

<html>
<body>
  <script src="https://js.puter.com/v2/"></script>
  <script>
    puter.ai.txt2img("A serene Japanese garden with cherry blossoms", {
      model: "stabilityai/stable-diffusion-3-medium"
    }).then(image => {
      document.body.appendChild(image);
    });
  </script>
</body>
</html>

We ran the image-generation version of the workload we use across our pricing guides: 500 monthly users generating 30 images each, for 15,000 images a month. On SD3.5 Medium through the API at $0.035 per image, that is about $525 a month, growing linearly with your user base (the same volume is roughly $450 on Core and $1,200 on Ultra). Through Puter.js the same app costs you $0 at any scale, because each user carries their own usage. Puter serves the open Stability models (stabilityai/stable-diffusion-3-medium and stabilityai/stable-diffusion-xl-base-1.0) rather than the hosted Core and Ultra services, so this is a fit when those models meet your quality needs.

Self-hosting under the Community License

If you generate under $1M in annual revenue, you can download SD3.5 and SDXL from Hugging Face and run them on your own hardware for free under the Community License. There is no per-image charge, only your compute cost, and your data never leaves your infrastructure. This is the most cost-effective route at high volume if you can run a GPU.

25 free signup credits

Creating a Stability account with Google's social login grants 25 free credits, enough for several Ultra generations or up to about eight Core images, to test the API before you buy. Signing up with a username and password does not include this credit.

Third-party trial credits

Stability models are also hosted on Replicate, Fal.ai, and AWS Bedrock, each of which has its own trial credit or free-usage terms. These are useful for evaluation, though at production volume you are back to paying per image.

Real-world cost examples

E-commerce product images

We modeled an online store with 500 products generating four lifestyle images each, for 2,000 images a month.

For catalog images where the product must look right, we would generate on Core for drafts and finalize the published shots on Ultra, landing between the two figures rather than paying Ultra rates on every attempt.

Daily marketing content

We calculated a content team producing 20 images a day, 600 a month, on Core: $18 in generation. Adding background removal (5 credits) and a Conservative upscale (40 credits) on the 100 finals each month adds about $45, for roughly $63 a month all in. The editing and upscaling steps cost more than the generation here, which is typical once a workflow has multiple stages.

In-app avatar generation

We modeled an app where 1,000 monthly active users each generate three avatars, for 3,000 images a month on Core: about $90. This is the workload where the User-Pays model changes the math, because moving generation to Puter.js shifts that $90 (and its growth as users increase) off your bill entirely.

Complete Stability AI API pricing table

All prices in credits; 1 credit = $0.01.

Generation

Edit and control

Upscale

Audio and 3D

Stability deprecated the SD1.6 API on July 24, 2025. For the full live catalog, see Stability's official pricing page.

Conclusion

Stability's flagship Stable Image Ultra costs $0.08 per image and Stable Image Core costs $0.03, with the SD3.5 family in between, all billed in credits at 1 credit = $0.01.

The main cost levers:

• Choose the cheapest model that meets your quality bar (Core is under half of Ultra).
• Self-host the open weights if you are under $1M in revenue.
• Use Fast or Conservative upscaling instead of Creative unless you need it.
• Reuse generated assets rather than regenerating.
• Ask about bulk pricing above roughly 10,000 images a month.

Related

• Free, Unlimited Stable Diffusion API
• How to Get a Stability AI API Key
• Free, Unlimited Image Generation API
• Free, Unlimited FLUX AI API
• FLUX API Pricing
• Free, Unlimited GPT Image API
• Free, Unlimited Imagen API
• Free, Unlimited Nano Banana API
• Ideogram API Pricing
• OpenAI API Pricing
• Gemini API Pricing
• Claude API Pricing
• Free, Unlimited AI API