Chinese AI company DeepSeek updated its V4.1 Flash-1017 model on October 17, and MiniMax also released its M3.1 Flash model (previously codenamed Space Bunny). Both Flash models performed excellently in evaluations. MiniMax M3.1 Flash scored 74 points in the DeepSWE test, while DeepSeek V4.1 Flash-1017 scored 77 points. MiniMax M3.1 Flash also achieved 76.2 points in the financial Agent evaluation BankerToolBench, tying with Claude Opus 5 for first place. MiniMax has halved the input and output prices of its new model compared to the previous generation M3, reduced cache read prices by approximately 75%, and achieved an output speed of 200+ TPS. These advancements indicate that Flash models are transitioning from being alternatives to flagship models to becoming a new competitive arena, challenging the capabilities of flagship models at a lower cost.