Unsloth has released version 3.1 of the Qwen3.8-Flash-Next GGUF models, claiming a 15% accuracy improvement over previous versions. The updated model files are available for download via the Unsloth documentation portal.