ukisai/Swift1.5-Qwen3.8-Flash-Next

UkisAI releases a derivative of Qwen3.8-Flash-Next that reduces thinking tokens by 63.4% and claims a 1.8x speedup. The model card documents the training approach and provides links to evaluation data. However, the source lacks explicit installation instructions or code snippets for loading the model, and the performance claims are not independently verified.