← all releases
Unsloth AI·May 13, 2026·v3.6·llm
Qwen3.6 GGUFs
Unsloth AI released experimental GGUFs for Qwen3.6 in 27B and 35B variants, featuring a new Multi-Token Prediction (MTP) architecture that delivers up to 220 tokens/s on a single GPU — a 1.4x speed-up over previous versions. The 27B model runs on 18GB RAM and the 35B-A3B on 22GB, making high-quality local inference more accessible.
LLMGGUFLocal InferenceQwenOpen SourceMulti-Token Prediction
overview
Unsloth AI released experimental GGUFs for Qwen3.6 in 27B and 35B variants, featuring a new Multi-Token Prediction (MTP) architecture that delivers up to 220 tokens/s on a single GPU — a 1.4x speed-up over previous versions. The 27B model runs on 18GB RAM and the 35B-A3B on 22GB, making high-quality local inference more accessible.
key features
- 01MTP (Multi-Token Prediction) architecture for up to 220 tokens/s on a single GPU
- 0227B model runs on 18GB RAM; 35B-A3B runs on 22GB
- 031.4x speed improvement over previous Qwen GGUF versions
related productsbrowse all →
Protenixopen sourceProtenix is an open-source, trainable PyTorch implementation of AlphaFold 3, designed for high-accuracy biomolecular structure prediction. It aims to advance accessible and extensible research tools for the computational biology community. ([github.com](https://github.com/bytedance/Protenix?utm_source=openai))Z-Imageopen sourceZ-Image is an efficient image generation foundation model developed by Tongyi-MAI, designed to produce high-quality, diverse, and stylistically versatile images. It serves as a robust base for creators, researchers, and developers seeking advanced image generation capabilities.