Ship custom models without the ops.
ModelForgeLab lets you version, share, and serve LoRA adapters and fine-tuned checkpoints — train new adapters from your own datasets, preview results live in the playground, and pull any artifact with a single CLI command.
No credit card. Free for teams up to 5 seats.
$ mfl pull sd15-pixelart-lora
✔ resolving adapter … 3.0.0 (safetensors, fp16)
✔ verifying sha256 … ok
✔ downloaded 36.0 MiB in 4.2s
$ mfl serve sd15-pixelart-lora --port 11434
▶ adapter ready at http://localhost:11434Adapter registry
Semantic versioning, sha256 integrity, and format-aware storage for safetensors, GGUF, and diffusers layouts.
Explore the registry →Fine-tune with LoRA
Upload a dataset, pick a base model, and queue a low-rank training run. Stream logs and pull the resulting adapter when it is done.
Start a run →Image & text preview
Sanity-check an adapter before you ship it. Generate a few images or completions and watch progress live.
Open the playground →Featured adapters
View allFLUX Game Concept Art
black-forest-labs/FLUX.1-dev
FLUX Hyperrealism
black-forest-labs/FLUX.1-dev
FLUX Anime Character XL
black-forest-labs/FLUX.1-dev
SDXL Anime Illustration
stabilityai/stable-diffusion-xl-base-1.0
SDXL Dark Fantasy
stabilityai/stable-diffusion-xl-base-1.0
Llama 3 8B Instruct (GGUF Q4)
meta-llama/Meta-Llama-3-8B-Instruct