<h3 id="about-these-variants">About these variants</h3>Four precision levels are available for MOSS-Image (available for both the SFT and Base checkpoints), trading VRAM and speed against generation quality: - BF16 — full-precision reference (~10 GB). Highest quality, largest footprint. Available as diffusers on the official <a rel="ugc" href="https://huggingface.co/OpenMOSS-Team/MOSS-Image-SFT/tree/main">OpenMOSS-Team/MOSS-Image-SFT</a> and <a rel="ugc" href="https://huggingface.co/OpenMOSS-Team/MOSS-Image-Base/tree/main">OpenMOSS-Team/MOSS-Image-Base </a>Hugging Face repositories, and ComfyUI compatible models on the <a rel="ugc" href="https://huggingface.co/Comfy-Org/MOSS-Image/tree/main/diffusion_models">Comfy-Org/MOSS-Image </a>Hugging Face repository.- INT8 (W8A8) ConvRot — 8-bit weights + 8-bit activations (~5.4 GB). Near-lossless quality (mean relative error ~0.87%, cosine similarity ~1.0005 vs BF16), ~2× smaller than BF16, runs on INT8 tensor cores (RTX 30-series and up).- INT6 (W6A8) — 6-bit weights + 8-bit activations (~4.1 GB). Middle-ground: smaller than INT8 with only a small quality trade-off.- INT4 (W4A8) — 4-bit weights + 8-bit activations (~3.0 GB). Smallest footprint; uses ConvRot with a per-tensor codebook that decodes to INT8 for compute, so it runs on the same INT8 hardware as W8A8. Larger quality trade-off than INT6. Two checkpoints available at each precision, same architecture, different intended use: - MOSS-Image-SFT (moss_image_sft_*) — supervised fine-tuned on top of Base. The default for day-to-day text-to-image generation, including natural-language prompts, tag sequences, and JSON layout control. - MOSS-Image-Base (moss_image_base_*) — foundation checkpoint for continued pretraining, fine-tuning, and LoRA training. Use this as the starting point when training adapters or doing further post-training; SFT is the better choice for end-user inference. MOSS-Image — INT8 / INT6 / INT4 ConvRot for ComfyUI Quantized weights of MOSS-Image for fast, low-VRAM inference in ComfyUI. These are modified (quantized) versions of the official MOSS-Image checkpoints from the OpenMOSS Team, repackaged for ComfyUI with mixed-precision quant metadata that loads via the native MOSS-Image ComfyUI nodes. The BF16 reference is linked for comparison but is available directly from OpenMOSS Team's official repositories. About MOSS-ImageA text-to-image diffusion model from the OpenMOSS Team, with a 5B DiT image transformer (42 layers, 252 GEMM linear projections) paired with a Qwen3.5-9B text encoder and a FLUX.2 small VAE. MOSS-Image supports natural-language prompts, tag-sequence prompts, and JSON-layout conditioning for precise text placement and object arrangement in the generated image. Highlights - Flexible Prompting — accepts free-form natural language, compact tag sequences, and structured JSON layouts in the same model, letting creators match the prompting style to the task. - Layout-to-Image Control — pass a JSON layout with a scene caption, items, bounding boxes (yxyx format, abs_centralized coordinates), and rendered-text specifications to control where elements appear in the frame. Key features - Unified Prompting Interface — natural-language, tag-sequence, and layout-JSON prompts all handled by one model, no mode switch. - Rendered Text and Layout — specify on-image text and element positions directly through JSON, useful for posters, signage, UI mockups, and composition-driven work. - Qwen3.5-9B Text Encoder — large language-model-grade text understanding for stronger prompt grounding than CLIP-class encoders. - FLUX.2 Small VAE — shared autoencoder path with the FLUX.2 ecosystem; VAE kept in BF16 across all quantized variants (quantizing a VAE this small saves little and risks reconstruction artifacts). - SFT and Base Variants — SFT for inference, Base for fine-tuning and LoRA training. Recommended inference settings (from the official MOSS-Image ComfyUI workflow):- Sampler: Euler- Scheduler: sgm_uniform- Steps: 50- CFG: 3- Shift: 5- Negative-branch skipping: layer 22- Resolution: 1024×1024 License MOSS-Image is released by the OpenMOSS Team under the <a target="_blank" rel="ugc" href="https://huggingface.co/OpenMOSS-Team/MOSS-Image-SFT/blob/main/LICENSE">[Apache 2.0 License]</a>. These files are a quantized derivative and are distributed under the same license — see the original repositories for full terms, permitted uses, and any commercial-use restrictions.

Moss Image INT8/INT6/INT4

mossImageINT8INT6_mossImageINT8.safetensors

mossImageINT8INT6_mossImageBaseINT8.safetensors

Moss Image Base INT8

Moss Image SFT INT4

Moss Image Base INT4

Moss Image SFT INT6

Moss Base Image INT6

Moss Image SFT INT8

MOSS Image is a 5B parameter text-to-image DiT from the OpenMOSS team, built for accurate in-image text, layout control and long prompts. It pairs a Qwen3.5 9B text encoder with the FLUX.2 VAE, accepts prompts of up to 4K tokens in natural language, booru-style tags or a JSON layout, and renders from 512px up to 2K at any aspect ratio.Originally released by <a target="_blank" rel="ugc" href="https://huggingface.co/OpenMOSS-Team">OpenMOSS</a> on Hugging Face. All credit for the model goes to the OpenMOSS team. Civitai hosts a mirror so creators can generate with it on-site and train against it. Head to the <a target="_blank" rel="ugc" href="https://github.com/OpenMOSS/MOSS-Image">original repo</a> for the reference pipeline, the user guide and updates.<h3>Two versions</h3><ul><li>SFT is the supervised fine-tune of Base on curated data. OpenMOSS recommends it for generation: cleaner text grouping and more organised layouts.</li><li>Base is the undistilled base model with no post-training. OpenMOSS recommends it for fine-tuning and LoRA training. LoRAs trained on Base can also be used on SFT, with partial compatibility.</li></ul>Both run at 50 steps with CFG 3 (raise it for short prompts), flow-matching Euler, shift 5. OpenMOSS recommends English prompts; Chinese and other languages are supported. Use line breaks and quotes around text you want rendered.<h3>Files</h3>Each version carries the diffusion model in bf16 and int8, from the <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/MOSS-Image">Comfy-Org repackage</a>. The Qwen3.5 text encoder and FLUX.2 VAE ship in the same repo. The original Diffusers weights are at <a target="_blank" rel="ugc" href="https://huggingface.co/OpenMOSS-Team/MOSS-Image-Base">MOSS-Image-Base</a> and <a target="_blank" rel="ugc" href="https://huggingface.co/OpenMOSS-Team/MOSS-Image-SFT">MOSS-Image-SFT</a>.<h3>Generating on Civitai</h3>Both versions are in the Civitai generator for text-to-image, with steps, CFG and a negative prompt exposed. JSON layout prompts work as plain prompt text. Image editing and reference images are not part of this release.<h3>License</h3>MOSS Image weights and inference code are released under the <a target="_blank" rel="ugc" href="https://huggingface.co/OpenMOSS-Team/MOSS-Image-Base/blob/main/LICENSE">Apache License 2.0</a>. Commercial use, derivatives and LoRAs are allowed; keep the licence notice with any redistribution.<h3>Links</h3><ul><li><a target="_blank" rel="ugc" href="https://github.com/OpenMOSS/MOSS-Image">GitHub: OpenMOSS/MOSS-Image</a></li><li><a target="_blank" rel="ugc" href="https://github.com/OpenMOSS/MOSS-Image/blob/main/USERGUIDE.md">User guide (prompting, layout JSON, fine-tuning)</a></li><li><a target="_blank" rel="ugc" href="https://huggingface.co/OpenMOSS-Team/MOSS-Image-SFT">MOSS-Image-SFT</a></li><li><a target="_blank" rel="ugc" href="https://huggingface.co/OpenMOSS-Team/MOSS-Image-Base">MOSS-Image-Base</a></li><li><a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/MOSS-Image">ComfyUI files</a></li></ul>

Files with SHA256: f601c1e1161915fb3fea003da0c1782e374dd960fbb3978bc0e4e3fa1f2d8e6f