I tested a free AI image model on my own GPU (+ an AI-built coding model)

This week in open AI: a coding model that an AI research system helped build, a free image model I tested step by step on my own RTX 4070 SUPER, a model you can chat with about DNA, and a diffusion LLM that writes 1,000+ tokens per second. Every source is on screen and listed below, and every local test is a real run. 0:00 Intro 0:30 Recursive · Flint 2:50 OpenMOSS · MOSS-Image (my local test in ComfyUI) 6:42 Hugging Face · Carbon-X 7:48 Inception · Mercury 3 8:40 Outro ▶ SOURCES (all benchmark numbers as reported by each company) Recursive Flint • Announcement: https://www.recursive.com/updates/int... • Model card: https://huggingface.co/RecursiveSI/flint • rec engine: https://github.com/recursive-oss/rec • Flint Arcade (community demo by alphaBell): https://huggingface.co/spaces/alphabe... MOSS-Image (OpenMOSS) • GitHub: https://github.com/OpenMOSS/MOSS-Image • Model card: https://huggingface.co/OpenMOSS-Team/... • Project page: https://openmoss.github.io/MOSS-Image/ • Technical report: https://arxiv.org/abs/2610.23374 ComfyUI (used for my test) • ComfyUI: https://github.com/Comfy-Org/ComfyUI • Model files for ComfyUI: https://huggingface.co/Comfy-Org/MOSS... • Official workflow template: https://github.com/Comfy-Org/workflow... Carbon-X (Hugging Face) • Model card: https://huggingface.co/HuggingFaceBio... • Code: https://github.com/huggingface/carbon Mercury 3 (Inception) • Announcement: https://www.inceptionlabs.ai/blog/int... Subscribe for the next brief:    / @openmodelbrief   #ai #opensource #comfyui