Will there be exmperimental 4gb VRAM support with smaller model quants?

[Marczellito]

I know i'm asking for a lot, don't treat it as a request feature, but i'm wondering - are there any plans for support of graphics cards with VRAM smaller than 8 GB? That would probably also require smaller quants of models, as even LFM2.5-24B-A2B takes about 4 GB of VRAM even before any expert.
Than you for your work!