OpenMOSS open-sources MOSS-Image: 5B DiT with CN/EN text rendering and JSON layout control
Xianbao_QIAN · x · 2026-10-17
OpenMOSS has open-sourced MOSS-Image on Hugging Face: a 5B diffusion transformer paired with a Qwen3.5 text encoder and FLUX.2 VAE, focused on Chinese & English text rendering and controllable layouts. Highlights: prompts up to 4K tokens with multi-line text, experimental JSON layout control via bounding boxes, multiple aspect ratios/resolutions, Diffusers/Gradio/ComfyUI workflows, and both Base & SFT checkpoints. A vLLM-Omni deployment recipe is included with Docker images (CUDA 13.0/12.9), single-GPU startup commands, and a /v1/images/generations API for serving either checkpoint.
Related event: Fudan's OpenMOSS Releases MOSS-Image, a 5B Open-Source Text-to-Image Model(5 posts)→
More from Infra
- OpenAI and Broadcom's ML accelerator Jalapeño makes Time's 2026 Top 100 Inventions list — thehiphopswami · 2026-10-17
- Broadcom's Ultra and Jericho Are the Underrated Half of the AI Ethernet Story — BenBajarin · 2026-10-17
- Analyst rebuts 'China floods HBM by 2028' claim, citing 5-7 year DRAM gap — JOBhakdi · 2026-10-17
- Bitter lesson for local LLMs: below 100 TPS, you won't actually use it — TheZachMueller · 2026-10-17
- AI is pulling quantum timelines forward, but hardware is at the vacuum-tube stage — robleclerc · 2026-10-17
- SentryWard: 32 of 85 sampled GPU kernels harbor race and sync bugs, formal verification for AI-generated code — JiaZhihao · 2026-10-17