MiniMax says M3.1 Flash is the model that ran anonymously as "Space Bunny" and processed more than 200 trillion tokens before the reveal. Pricing is flat: $0.15 per million input tokens, $0.60 per million output and $0.015 for cached input. For comparison, MiniMax-M3 on the standard tier costs $0.30 and $1.20 for prompts up to 512K tokens, per MiniMax's pricing page.
The company also says inference is nearly three times faster, with output at 200+ tokens per second and a median of 170+. It claims results on par with top models on Automation-Bench, DeepSWE and Agents' Last Exam. Those are MiniMax's own numbers, and I haven't seen independent runs yet.
It's available through the MiniMax API, M Plan and the MiniMax Code app. The API takes OpenAI-style requests at https://api.minimax.io/v1 and Anthropic-style requests at https://api.minimax.io/anthropic, with the model ID MiniMax-M3.1-Flash. Thinking is on by default, and reasoning effort defaults to max (it accepts low, medium, high, xhigh and max). MiniMax says the weights come later, and so does a Pro model.