MiniMax Launches M3.1 Flash AI Model With Nearly 3x Faster Inference
Chinese AI company MiniMax has launched M3.1 Flash, an updated AI model designed to improve coding, computer use, cybersecurity and autonomous agent tasks while reducing the cost of running AI-powered workflows. The company says the model’s online inference speed has nearly tripled compared with its predecessor, while usage costs for agentic tasks have fallen by 70%.
Announced on October 19, M3.1 Flash targets developers and businesses that need AI systems to complete multi-step tasks quickly and affordably. The release reflects growing competition among AI companies to make models more practical for everyday software development and automated business operations.
Faster AI Agents and More Efficient Workflows
AI agents are designed to do more than respond to individual prompts. They can plan tasks, use external tools, interact with software and carry out sequences of actions toward a defined goal. However, these workflows can require many model calls, making response time and operating costs important considerations.
MiniMax says M3.1 Flash improves performance across complex agentic tasks, programming, computer use and cybersecurity. Its published results include a 73.7% score on the DeepSWE v1.1 coding benchmark, 61.7 points on OSWorld V2 for computer-use tasks, and an 82% score on CyberGym for vulnerability discovery and validation.
The company also reports a score of 76.2 on BankerToolBench, which evaluates complex financial workflows. These results are part of MiniMax’s own launch disclosures and should be considered company-reported performance figures rather than independent confirmation that the model outperforms competing systems in everyday use.
Lower Costs Could Broaden AI Adoption
MiniMax lists API pricing at $0.15 per million input tokens and $0.60 per million output tokens, with cached input priced at $0.015 per million tokens. It says the overall cost of agentic tasks is 70% lower than with its predecessor.
Lower prices and faster inference could make it easier for businesses to run AI agents repeatedly across coding, customer support, financial operations and other workflows. The practical benefit will depend on how reliably the model completes tasks, how much human oversight is required and whether the claimed savings hold across different workloads.
M3.1 Flash is available through the MiniMax API, M Plan and MiniMax Code. The company says model weights will be released later, meaning developers cannot yet treat the announcement as a confirmed open-weight release.
Competition Is Shifting Beyond Model Capability
As AI systems become more capable, speed, cost and reliability are increasingly important factors in choosing a model. A system that performs well but requires expensive or lengthy inference may be less practical for high-volume automated work than a more efficient alternative.
MiniMax’s announcement positions M3.1 Flash around this balance of capability and cost. However, independent benchmark comparisons and real-world testing will be important for determining how its improvements compare with rival models and whether the reported gains translate into measurable productivity.
For now, the launch adds another contender to the race for faster, lower-cost AI agents. Its longer-term significance will depend on how the model performs outside the company’s own evaluations and whether developers see consistent benefits when deploying it at scale.
