ParentFull threadsatvikpendem·Use the MTP models which 2x token generation speed, for example: https://unsloth.ai/docs/models/qwen3.6#mtp-guideView on HN