Qwen3.8-Flash-Next on 12GB VRAM - 65 tokens per second

A developer reported significant performance improvements for the Qwen3.8-Flash-Next model on a 12GB RTX 5070. Initially achieving 15 tok/s output and 100-120 tok/s prompt processing with IQ3_XXS quant using llama.cpp, they later developed a custom inference engine. This engine boosted performance to ~65 tok/s output and ~430 tok/s prompt processing with the same IQ3_XXS quant, with 2-bit quants using RCO-GSQ quantization running even faster.