Stability AI iconStability AI

Maker of Stable Diffusion 3.5, Stable Audio 3.0, Stable Video and Stable Fast 3D: open-weight models plus a credit-based API.

27.3k+ GitHub Stars
$0.01 Per API Credit
$1M Free License Revenue Cap

Overview

Stability AI is the London company behind Stable Diffusion, the open image model it first released in August 2022. Today it makes models for four media: images (Stable Diffusion 3.5 Large, Large Turbo and Medium, plus SDXL), video (Stable Video Diffusion), 3D (Stable Fast 3D, SPAR3D, Stable Video 4D) and audio (Stable Audio 3.0, released in May 2026 and trained on fully licensed data). You can use them three ways: download the open weights from Hugging Face and run them yourself, call them through the Stability AI Platform API at $0.01 per credit, or reach them through cloud partners such as AWS. The Community License makes the weights free for research and for businesses under $1M in annual revenue; bigger companies need an Enterprise License. Since Prem Akkaraju became CEO in June 2024 the company has leaned into entertainment: a co-development deal with Electronic Arts, music partners Universal and Warner, and a $76M Series B in August 2026 backed by EA, Sony Music, UMG and Warner Music. Also from Stability AI: Stable LM, its older open language models, now dormant.

The Verdict

Who Should Use Stability AI?

Best For

  • Developers who want image, audio and 3D generation from one API with clear per-call credit prices
  • Teams that need to self-host and fine-tune open weights (LoRAs and fine-tunes are allowed under the Community License)
  • Small studios and indie creators under $1M revenue who want free commercial use of the weights
  • Musicians and media companies that want an audio model trained on licensed data
  • Game and 3D teams that need fast single-image-to-3D assets
  • ComfyUI users who rely on the large Stable Diffusion tooling ecosystem

Not Ideal For

  • Casual users who want a polished chat-style image app (a consumer tool like Midjourney is simpler)
  • Video-first creators: the only video model listed is Stable Video Diffusion, with short clips of 14 or 25 frames
  • Companies over $1M revenue that want to self-host without negotiating an Enterprise License
  • Buyers who need a fixed monthly plan for the API (it is pay-as-you-go credits)

What's Great

  • Open weights across image, audio and 3D, with free commercial use under $1M revenue
  • You own your outputs, and fine-tunes and LoRAs are explicitly allowed
  • Transparent API pricing per call, from 2.5 credits (SD 3.5 Flash) to 8 credits (Stable Image Ultra)
  • SD 3.5 Medium runs on consumer GPUs (9.9 GB VRAM, excluding text encoders)
  • Stable Audio 3.0 is trained on fully licensed data and makes tracks of over six minutes
  • Stable Fast 3D turns one image into a 3D asset in about half a second
  • Full editing toolkit in the API: inpaint, outpaint, erase, relight, sketch and structure control, upscaling

Watch Out For

  • Only 25 free API credits for new accounts, then you pay as you go
  • Enterprise License needed above $1M revenue, with custom pricing only
  • The Community License is revocable if you break its terms
  • Video lags the field: no newer video model than Stable Video Diffusion, and no video entry in the API price list
  • The biggest audio model (Stable Audio 3.0 Large) is not open weights; it is API and enterprise only
  • Copyright litigation is still a factor: the UK Getty case went mostly Stability's way in November 2025, but Getty won a narrow trademark point on older models and an appeal is likely
  • The main generative-models repo has not been pushed since December 2025

Pricing

View all features & details

Image Models

  • Stable Diffusion 3.5 Large: 8.1B parameters, MMDiT architecture (Oct 2024)
  • Stable Diffusion 3.5 Large Turbo: distilled, 4-step generation
  • Stable Diffusion 3.5 Medium: 2.5B parameters, 0.25 to 2 megapixels
  • Stable Diffusion XL (3.5B parameters) and SDXL Turbo (one-step)
  • API-only services: Stable Image Ultra, Stable Image Core, SD 3.5 Flash

Video, 3D and Audio

  • Stable Video Diffusion: 14 or 25 frames at 3 to 30 fps
  • Stable Fast 3D: single image to 3D asset in about 0.5 seconds
  • SPAR3D: single-image 3D in under a second, with editing
  • Stable Video 4D, SV3D, Stable Zero123, TripoSR (with Tripo AI)
  • Stable Audio 3.0 Small SFX, Small (up to 2 min), Medium (up to 6:20), Large (over 6 min)
  • Stable Audio DAW plugin and StableAudio.com app (Aug 2026)

API Tools

  • Edit: erase object, inpaint, outpaint, remove background, search and replace, search and recolor, replace background and relight
  • Control: sketch, structure, style guide, style transfer
  • Upscale: fast, conservative and creative upscalers
  • Also on cloud partners, including Amazon Bedrock

License Terms

  • Community License covers SD 3.5, SDXL Turbo, Stable Fast 3D and Stable Audio 3.0, among other core models
  • Free commercial use under USD $1M annual revenue
  • You own outputs, subject to law and the Acceptable Use Policy
  • Fine-tunes and LoRAs allowed; new architectures derived from core models are restricted

Company & Funding

Company Background

  • Founded 2019 in London by Emad Mostaque and Cyrus Hodes
  • Mostaque stepped down as CEO in March 2024
  • CEO Prem Akkaraju (ex-Weta Digital) since June 2024
  • Board includes James Cameron, Sean Parker and Coatue's Thomas Laffont

Funding & Partners

  • $76M Series B (Aug 2026), $232M raised in total
  • New investors: Electronic Arts, Sony Music Group, Universal Music Group, Warner Music Group, AMD Ventures
  • EA deal (Oct 2025) to build AI tools for game textures and 3D previs
  • Technology partners include AWS, Microsoft Azure, NVIDIA, Arm and Lenovo

How It Compares

Feature Stability AI FLUX (Black Forest Labs) Midjourney Adobe Firefly
Open weights Image, audio, 3D Some image models No No
Self-hosting Yes, Community License Open models only No No
Media covered Image, video, 3D, audio Image Image-first Image, vector, in Adobe apps
Public API Yes, credit-based Yes No See Adobe
Best for Developers and self-hosters Image quality via API Artistic style Adobe users

Sources: Stable Image · SD 3.5 Large card · API pricing · see also FLUX, Midjourney, Adobe Firefly and Civitai (community models built on Stable Diffusion)

Is this your tool? Add a badge