Model reference · open weights
LongCat-Image-Dev is an open-weight image model from meituan-longcat. LongCat-Image-Dev (BF16) weighs 29.3 GB; the smallest configuration that runs it is RTX 4090 24 GB.
For self-hosting, compare this build's requirements below with the GPU you plan to rent.
Summary of the meituan-longcat/LongCat-Image-Dev model card, 2026-10-01
What it is
| Released by | meituan-longcat |
|---|---|
| Released | 2025-12-04 |
| VRAM | 29.3 GB for the weights |
What it runs on
| Card | One image | Memory |
|---|---|---|
| RTX 3060 12 GB … RTX 4060 Ti 16 GB | does not fit | |
| RTX 3090 24 GB | tight (encoders offloaded) | 23.4 GB |
| RTX 4090 24 GB | tight (encoders offloaded) | 23.4 GB |
| RTX 5090 32 GB | fits (encoders offloaded) | 31.0 GB |
| L40S 48 GB | fits | 44.0 GB |
| A100 80 GB | fits | 78.2 GB |
| H100 80 GB | fits | 78.1 GB |
| RTX PRO 6000 Blackwell 96 GB | fits | 93.8 GB |
| DGX Spark (GB10) 128 GB unified | fits | 107 GB |
| H200 141 GB | fits | 138 GB |
| B200 180 GB | fits | 176 GB |
From the model card
LongCat-Image-Dev is a development variant of LongCat-Image, representing a mid-training checkpoint that is released to facilitate downstream development by the community, such as secondary fine-tuning via SFT, LoRA, and other customization methods.
🔧 True Developer-Ready Foundation: Unlike typical release-the-final-model-only approaches, we provide the Dev—a high-plasticity, unconstrained state that avoids RL-induced rigidity. This enables seamless fine-tuning without fighting against over-aligned parameter spaces.
🛠️ Full-Stack Training Framework: We ship production-ready code for SFT, LoRA fine-tuning, DPO/GRPO/MPO alignment, and specialized Edit training. Every stage from pre-training data curation to reward model integration is reproducible, empowering researchers to build on our exact pipeline rather than reverse-engineering it.
Quoted from the model card on Hugging Face. The full card is behind the Hugging Face link above.
Running it yourself
Rent a machine by the hour. Runs as it is with diffusers, on the machine, in Python.
# on your rented machine: pip install diffusers transformers accelerate
import torch
from diffusers import DiffusionPipeline
pipe = DiffusionPipeline.from_pretrained("meituan-longcat/LongCat-Image-Dev", torch_dtype=torch.bfloat16).to("cuda")
image = pipe(prompt="a red bicycle on a cobbled street").images[0]
image.save("/workspace/out.png")