Efe

Local LLM inference on consumer hardware — VRAM requirements, KV cache sizing, quantisation trade-offs, and mixture-of-experts memory behaviour. I publish the underlying datasets openly under CC BY 4.0.