OPEN LINK ↗
// 1h agoOPENSOURCE RELEASE
ComfyUI 0.42.0 adds native MOSS-Image support
ComfyUI 0.42.0, released October 18, makes MOSS-Image a first-class workflow with native loaders, layout prompting, INT8 support, and fused inference optimizations. MOSS-Image is a 5B Apache-2.0 text-to-image model focused on layout control and accurate text rendering. [Official release](https://github.com/Comfy-Org/ComfyUI/releases/tag/v0.42.0) [MOSS-Image repository](https://github.com/OpenMOSS/MOSS-Image)
// ANALYSIS
This is meaningful infrastructure work: ComfyUI is turning a fast-moving community model into a standardized, reproducible local workflow.
- –Native support removes the custom-node friction around MOSS-Image checkpoint loading and layout prompts. [Implementation PR](https://github.com/Comfy-Org/ComfyUI/pull/17095)
- –Fused `situ` support targets roughly 30% faster INT8 inference, though gains remain hardware-dependent. [Optimization PR](https://github.com/Comfy-Org/ComfyUI/pull/17099)
- –Fused q/k/v detection and LoRA mapping make existing MOSS-Image adapters easier to reuse.
- –Developers should still expect substantial memory and compute demands: the model uses a Qwen3.5-9B text encoder and defaults to 50 sampling steps.
// TAGS
comfyuiopen-sourceimage-geninferencequantizationdevtool
DISCOVERED
1h ago
2026-10-19
PUBLISHED
1h ago
2026-10-19
RELEVANCE
9/ 10
AUTHOR
TraffAlex
