MiniMax has launched MiniMax M3.1 Flash, a new model it describes as a generational leap for agentic AI work, coding, computer use, and cybersecurity. Alongside the release, the company finally confirmed what many had suspected: the anonymous “Space Bunny” model that quietly topped leaderboards while processing more than 200 trillion tokens was an early version of M3.1 Flash.
According to the company, M3.1 Flash delivers online inference nearly three times faster than its predecessor while cutting usage costs by 70%. API pricing is set at $0.15 per million input tokens and $0.60 per million output tokens. The model is available through the MiniMax API, the M Plan subscription, and MiniMax Code for desktop.
The stealth run proved its value. OpenRouter, which hosted Space Bunny, said it was the platform’s number-one model for 10 straight days, and that real user feedback during the anonymous period helped MiniMax fix bugs and ship upgrades: time to first token improved 38%, throughput rose 45%, and the error rate fell from 1.4% to 0.7%.
On the benchmark front, MiniMax reports 73.7% on DeepSWE v1.1 for software engineering, 76.2 on BankerToolBench for enterprise workflows, and 82% on CyberGym for cybersecurity tasks. The company also says the model’s open weights will arrive on October 23, 2026.
The launch intensifies competition among frontier models. See how other leading systems compare in our DeepSeek, ChatGPT, and Gemini profiles, or read our ChatGPT vs Claude comparison for a head-to-head breakdown.
Source: Unite.AI
