view post Post 4,951 Weβre releasing new Qwen3.6 quants that run 2.5Γ faster on your GPU. β‘Qwen3.6-27B NVFP4 runs on 24GB VRAM.35B-A3B can hit 17,561 tok/s (B200).We also improved accuracy, tool calling, agent use, and looping.Qwen3.6 NVFP4: https://huggingface.co/collections/unsloth/nvfp4Guide: https://unsloth.ai/docs/models/qwen3.6#nvfp4 See translation 1 reply Β· π 15 15 π₯ 11 11 π€ 1 1 + Reply