Tongyi Qianwen has publicly thanked the Unsloth team for releasing the Qwen3.8-27B GGUF quantized version, describing it as smaller and smarter. The new Unsloth Dynamic V3 3.0 quantization solution boosts accuracy by over 10% compared to similar solutions at the same size, delivering performance close to full precision in key tasks, while its 1-bit variant runs on 8GB memory. Ultra-compressed versions cut file sizes by around 89% relative to full precision. These advancements lower the threshold for local deployment of large models, enabling regular devices to use the model with its multi-modal capabilities and Apache 2.0 license, reducing reliance on cloud services. Developers can download the files from Hugging Face following the provided running guides.