[Benchmark] Optimal DFlash2 quants for speed and context size, 5090 RTX, llama.cpp, Qwen 3.8 27B Dynamic3 Unsloth. Comparison with MTP

Graph: A cumulative metric of Speed x Context Size - the higher the dot - the better. Helpful for looking for the most balanced solution. TLDR: Q4 quants for DFlash2 demonstrate the highest speeds due to their higher acceptance rate. Q2 free up so much context that they are comparable with MTP in th