MiniMax Launches M3.1 Flash: Nearly 3x Faster Inference, 70% Lower Cost

What happened

MiniMax announced on its official blog that MiniMax M3.1 Flash is officially rolling out today. The company describes it as a next-generation Flash model with a generational leap in agentic tasks, coding, Computer Use and cybersecurity; everyday output speeds can reach 200+ tokens per second, with a median of 170+ TPS.

MiniMax also disclosed that 'Space Bunny' — the anonymous model that processed more than 200 trillion tokens — was an early version of M3.1 Flash. The model is now available through the MiniMax API, M Plan and MiniMax Code, with model weights to be released later; the company also teased a forthcoming Pro version focused on pushing the boundaries of intelligence. Reuters reported the launch the same day.

Key facts

1. Launch: October 19, 2026, on MiniMax's official blog
2. Positioning: A next-generation Flash model built for agentic tasks at high cost efficiency
3. Speed: Online inference nearly 3x faster; everyday output of 200+ TPS, median 170+ TPS
4. Pricing: Per 1M tokens: $0.15 input, $0.60 output, $0.015 cached input; the company says overall agentic-task cost is 70% lower than its predecessor
5. Benchmarks: Among top performers on AutomationBench-AA; 76.2 on BankerToolBench; 73.7% on DeepSWE v1.1; 61.7 on OSWorld V2 and 29.6 on Agents' Last Exam (first among Chinese models on both, per the company); 82% on CyberGym
6. Availability: Available now via the MiniMax API, M Plan and MiniMax Code; model weights to be released later