Local LLM Frontier

What is the best language model your GPU can run? Pick your VRAM budget, context length and KV cache precision: the chart shows which open-weight models fit at which quantization, how good they are on a shared intelligence index, and which closed frontier model that roughly corresponds to.