MiniMax released M3.1 Flash. It improves agentic tasks, coding, computer use and cybersecurity. Online inference is nearly three times faster. Usage costs drop by 70%. It scores 76.2 on BankerToolBench. It scores 73.7% on DeepSWE v1.1. It scores 61.7 on OSWorld V2. It scores 29.6 on Agents’ Last Exam. It scores 82% on CyberGym. Output speeds reach 200+ tokens per second. Flat pricing is $0.15 input, $0.60 output, $0.015 cached input per 1M tokens. It is available via MiniMax API, M Plan and MiniMax Code. Model weights will be released later.
What changed
M3 Flash had higher costs and slower inference speeds.
What it unlocks
Running agentic tasks, coding, computer use and cybersecurity workflows at lower cost and higher speed.
- 70% lower usage cost
- $0.15 per 1M input tokens
- $0.60 per 1M output tokens
- $0.015 per 1M cached input tokens
What you need to act on it
- access via MiniMax API, M Plan, or MiniMax Code
Sources