16GB for the full model, but the quantized version runs on ~3GB

→ View original post on X — @sumanth_077,