Unsloth vs bartowski MTP ggufs performance comparison

View on Reddit

A Reddit user benchmarks unsloth and bartowski MTP (Multi-Token Prediction) GGUF quants for Qwen3.5-4B. Bartowski uses Q8_0 for the MTP head, leading to larger file sizes but potentially different decoding speed.