Available Models
Sogni Studio ships with a large catalog of models you can run on-device or via the Sogni Supernet. The Model Explorer (sidebar → switch-arrow icon) is the source of truth — this page covers the headline groups; open the Explorer in-app for the full live list.
#Featured models
The current Sogni Studio defaults and most-used models:
| Model | Use for | Where it runs |
|---|---|---|
| Flux.2 [dev] | Top-tier prompt adherence, scene complexity, context conditioning | Supernet (Premium Spark) |
| Flux.1 Krea [dev] | Photorealism, fine-grained scenes | Supernet (Premium Spark) |
| Flux.1 [schnell] | Fast Flux at 1–4 steps | Supernet |
| Flux.1 Kontext [dev] | Prompt-based image editing | Supernet |
| Z-Image Turbo | Fast iteration, good defaults | On-device + Supernet |
| Qwen Image Edit 2511 | Vision-language editing, preserve identity | Supernet |
| Chroma v.46 / v.48 | Stylized output, color-rich aesthetic | Supernet |
| 🅂 Sogni.XL 𝛂2 / Artist / Photo | First-party Sogni-tuned looks | On-device + Supernet |
#Premium Spark models
These models require a Premium Spark balance (purchased by credit card or App Store IAP), separate from your free Spark:
- Flux.2 [dev]
- Flux.1 Krea [dev]
- OpenAI GPT Image 2
- ByteDance Seedance 2.0
OpenAI GPT Image 2 and ByteDance Seedance 2.0 are hosted closed-weight models — they appear in the Model Explorer when you're on Fast Supernet with Premium Spark available.
#Video models
- LTX-2.3 — image-to-video
- LTX-2.3 Dev — text-to-video
- Wan 2.2 — first-party Sogni video model
- Distilled variants of each for faster turnaround
See Creating Videos for which workflow each model supports.
#Audio models
- ACE-Step Turbo / SFT — music generation, full-song with lyrics or instrumental. See Creating Audio.
#Stable Diffusion lineage
The full SD lineage is browsable in the Model Explorer. Highlights:
SDXL / SDXL-Turbo / SDXL-Lightning families: SDXL v1.0, SDXL v1.0 with Refiner, Juggernaut XL v6, AlbedoBase XL v1.2 / v2.1, DreamShaper XL, DynaVisionXL, FaeTastic XL v2, LEOSAM's HelloWorld XL v5/v7, NightVision XL v9, Pony Diffusion XL v6, ProtoVision XL, Realistic Vision v5.1, ThinkDiffusion XL, ZavyChromaXL v4/v7/v8, plus all the SDXL-Turbo and SDXL-Lightning fast variants.
SD3: Stable Diffusion 3 with its triple-encoder MMDiT denoiser, available where Supernet workers support it.
SD 1.x / 2.x community models: A-Zovya RPG Artist Tools v3/v4, AbsoluteReality v1.8.1, Abyss OrangeMix v3, AnimeLineart, Architecture RealMix v1.1, Blazing Drive v13md, Blue Pencil XL, ChilloutMix, Children's Stories v1A, Crystal Clear XL, CyberRealistic family (Classic, Pony, v3.2, v5, v5 [6bit], XL v2), Deliberate v6, Dream Shaper v8.0, EdgeOfRealism v1/v2, EpiCRealism Pure Evolution v3/v5, epiCRealism XL, GhostMix v2, 国风 GuoFeng3/4/RealMix, InsanelyXL v1, InkArchitecture, Juggernaut Reborn, MagicMix + Realistic Vision, Mo Di Diffusion v1, Omnium XL v1.1, Openjourney v4.0, PicX Real v1, Pixel Alchemy Hyper v2.2, Playground v2, RealCartoon-Realistic v16, RealDream v14, Realistic Vision v5.1, Segmind SSD-1B, SoteMix, Stable Diffusion v1.4/1.5/2.0/2.1, Sweet Mix v2.1/v2.2, The Araminta Experiment CV6, Vivid OrangeMix, Yamer's Realistic XL5, YesMix v2/v4/v5 — and more.
LCM (Latent Consistency Models): LCM versions of Sogni, Blue Pencil XL, Comic Craft, CyberRealistic, Delusions v2, DreamCrafter v1, DreamShaper v7/v8, MagicMix + Realistic Vision, RealDream v7, SD v1.5, SoteMix. Use the LCM scheduler for 2–8 step generation with low guidance (1–2).
#On-device quantized variants
Many models are available in [6-bit], [4.5-bit], and [4-bit] flavors — ~3× lighter than the originals — for running on Macs with less unified memory. The full-precision versions are the default where there's headroom.
#Importing more
Anything in SafeTensors or CKPT format can be converted and added to your local library. See Importing Stable Diffusion Models.
Need help? Ask in our Discord!
For questions about the model lineup, requests for new models, or tips — join the Sogni Discord ✨