MiniMax officially rolled out MiniMax M3.1 Flash on October 19, 2026. The announcement page carries a date without a time of day. The company identifies Space Bunny — the model that processed more than 200 trillion tokens while operating anonymously — as an early version of M3.1 Flash. Read the announcement.
Positioning and speed
As the next-generation Flash model, M3.1 Flash focuses on agentic tasks, coding, computer use and cybersecurity. MiniMax reports that it makes online inference nearly three times faster and reduces usage costs by 70 percent compared with the previous generation.
Vendor-reported benchmarks
- Automation: ranks among the top performers on AutomationBench-AA, completing real enterprise workflows across SaaS systems; scores 76.2 on BankerToolBench for multi-step planning, document consistency, financial compliance and chained tool calls.
- Coding: 73.7% on DeepSWE v1.1, which requires models to understand code, develop features, fix issues and validate their work through testing.
- Computer use: 61.7 on OSWorld V2 and 29.6 on Agents' Last Exam, which MiniMax says ranks first among Chinese models on both benchmarks.
- Cybersecurity: 82% on CyberGym, with internally evaluated strength in code auditing, vulnerability patching, penetration testing, malware analysis and reverse engineering.
These are vendor-reported results; no independent evaluation accompanied the announcement. Availability and pricing details follow MiniMax's official platform pages.
For how we treat vendor benchmarks, see our verification guide.