MiniMax claims the Space Bunny model as M3.1 Flash and moves plan users to it

MiniMax has started moving developers onto its next model. Chinese business outlet Yicai reported on October 19, 2026 that MiniMax's official account posted a short video that afternoon captioned 'Welcome back, Space Bunny', which was read as the company claiming the anonymous Space Bunny model. According to Yicai, Space Bunny appeared in late September and within three days reached the top of daily token usage on OpenRouter and OpenCode, sharing the top three with DeepSeek V4.1 Flash and GLM 5.3 Flash. Yicai says it is an early version of MiniMax M3.1 Flash, which is due for a formal launch soon.

MiniMax's own API model list already shows MiniMax-M3.1-Flash, described as a multimodal coding model with a 1M token context window and adjustable thinking depth. The same page sets out the retirement of the M2.5 and M2.7 families: from October 19 they are closed to new users, and on November 18, 2026 they stop working for pay-as-you-go API customers.

For M Plan and Token Plan subscribers, MiniMax says requests to M2.5 and M2.7 are routed automatically to M3.1 Flash starting October 19. Old model names keep working for compatibility and plan allowances stay the same, but MiniMax asks developers to check their apps and plan the move.

Why it matters

Teams building on MiniMax's low-cost coding models are now running on a new model behind the same names, so outputs can change even if they changed nothing.

Source: MiniMax