Add some lite version for 6gb vram device

Adding support for 6 GB devices would broaden accessibility, allow more users to test the platform, and encourage community adoption. The lite mode could include: Smaller/quantized model options (e.g., 3B–7B Q4 instead of 13B full precision). Automatic offloading to system RAM when VRAM runs short. A fallback UI experience that disables heavy features but keeps core chat functional. This would not only expand the supported hardware base but also provide a smoother on-ramp for users who want to try the product before upgrading their GPU.