Stable Diffusion

Editor's Review

Stable Diffusion in 2026 is the open-weights veteran in a strange spot: there is still no Stable Diffusion 4 as of October 2026 — the SD 3.5 family (October 2024) remains the flagship, scoring 92% on independent visual-fidelity tests — while the company pivoted hard to AI music. Under CEO Prem Akkaraju (with Napster's Sean Parker advising), Stability refocused on professional AI music tools: three audio models plus music-editing software, with Sony, Warner, and Universal investing in the August 2026 round and licensing catalogs for training. Licensed, commercially-safe generative audio is the new strategy. Pricing: self-hosted is $0 — open weights downloadable (Hugging Face), run locally via Automatic1111/ComfyUI/Forge, free under the Community License for individuals and businesses under $1M annual revenue (enterprise license above that). The Developer API is pay-as-you-go at 1 credit = $0.01: Stable Image Ultra 8 credits ($0.08), SD3.5 Large 6.5 ($0.065), Large Turbo 4 ($0.04), Medium 3.5 ($0.035), Flash 2.5 ($0.025), SDXL from 0.9 ($0.009); 25 free credits to start. Stable Assistant (consumer chatbot on SD 3.5): $9–99/month tiers. The differentiator hasn't changed: no ceiling, fully private local runs, fine-tune anything. The honest costs: self-hosting needs GPU hardware and technical wrangling; first-time outputs can look plain next to Midjourney; the $1M revenue gate trips up mid-size companies; and open licensing is what made a sustainable paid business so hard in the first place. Who should pay? Self-hosters, fine-tuners, and privacy absolutists. Versus FLUX 3 (the newer open-weight-descended frontier, pay-per-image from $0.024): Flux wins current quality; SD wins ecosystem maturity. Versus Midjourney ($30/month): Midjourney wins ease. SD is the infrastructure pick.