Space Bunny (anonymous M3.1 Flash) topped OpenCode for 12 days
OpenCode confirms Space Bunny was M3.1 Flash: it led the board 12 straight days, served 106T tokens by Oct 5, about 1.7x DeepSeek V4.1 Flash, and is free for three days.
Space Bunny has a name. @OpenCode has confirmed it was @MiniMax M3.1 Flash. For twelve consecutive days it held the top of the @OpenCode model board. In the window to 5 October it served 106 trillion tokens, about 8.8 trillion a day. The runner-up, @DeepSeek V4.1 Flash, served 62 trillion over the same period. Space Bunny did roughly 1.7 times the volume. That is not a trial run. That is a model people kept choosing. The interesting part is not the reveal. It is the method. For twelve days, developers picked a model with no brand, no benchmark post and no vendor claim attached. They picked it on output alone. It won that test before anyone knew whose it was. A blind taste test at production scale is a better signal than any launch deck. @MiniMax now has a named product and a distribution channel in a single move. Space Bunny is on @OpenCode from today, free for three days. The chart also says something quieter. Six of the ten most used models on that board are from Chinese labs, and most of the top ten are open-weight. The default stack is drifting. If you have a workload that suits a fast flash model, the next three days are the cheapest test you will get. The board is public. The next stealth model will be watched the same way. #OpenCode #MiniMax #AIModels #OpenWeights