Locally Uncensored v2.3.0 adds hardware-aware bundles

This is the kind of update that matters for local AI adoption: less model-hunting, less setup friction, and more “install, detect, recommend, run.” The product is moving from a pile of features toward a genuinely turnkey desktop runtime for local chat, image, and video.

• –VRAM-aware onboarding is the strongest feature here; it directly addresses the most common local-AI failure mode: users picking models their GPU cannot handle.
• –One-click bundles and ComfyUI auto-install make the app feel closer to an appliance than a toolkit, which should help non-expert users stick with local workflows.
• –Adding GLM 5.1, Qwen 3.5, and Gemma 4 keeps the app aligned with current open-weight model momentum instead of freezing on older defaults.
• –FramePack I2V on 6GB VRAM is notable because local video generation has usually been gated by painful memory requirements.
• –No Docker and Windows/Linux support widen the addressable audience for people who want local inference without infra overhead.