Space Bunny (anonymous M3.1 Flash) topped OpenRouter for 10 days on 56.7T tokens
The stealth model held #1 on OpenRouter's coding board for ten days and burned 56.7T tokens; it is MiniMax M3.1 Flash at a third of M3's cost, 3x faster, $0.15/M input, 70% cheaper per task, weights on Oct 23.
for 10 days straight the best coding model on OpenRouter was a stealth bunny nobody could identify. it just sat at #1 and quietly ate 56.7 trillion tokens. turns out it was MiniMax M3.1 Flash the whole time. and it runs at a third of the cost of M3, 3x faster. you were probably already losing to it without knowing its name. frontier coding and agent scores, $0.15 per million tokens in, 70% cheaper per task than the last model. the only number that matters for agents is cost per finished task. this is the first flash model that makes that math a joke. weights drop oct 23.
This post quotes another post: x.com/minimax_ai/status/2112150867868541221