Key Notes
- MiniMax M3.1 Flash is now available through the company’s API, M Plan and MiniMax Code, targeting agents, coding, computer use and cybersecurity workflows.
- The API is priced at $0.15 per million input tokens, $0.60 per million output tokens and $0.015 per million cached input tokens.
- MiniMax says M3.1 Flash reaches more than 200 output tokens per second in everyday use; weights and a more capable Pro model are planned for later release.
MiniMax has launched M3.1 Flash, a new model aimed at AI agents that write code, operate software and carry out multistep tasks. The company is positioning the release around a familiar trade-off for agent builders: capable models can become expensive and slow once they use tools, retry steps and keep track of long-running work. M3.1 Flash is MiniMax’s attempt to push that balance toward lower-cost, higher-throughput inference.
In its announcement, MiniMax said the model is now available through its API, M Plan and MiniMax Code. API pricing is flat at $0.15 per million input tokens, $0.60 per million output tokens and $0.015 per million cached input tokens. The company says this brings the total cost of agentic workloads down 70% from its predecessor, though real costs will still depend on prompt size, tool calls, retries and the amount of output an application needs.
Built for Long-Running Agent Work
MiniMax is targeting work that requires more than a single answer: automated tasks across SaaS tools, coding projects, computer use and security analysis. The company says M3.1 Flash reaches the level of leading systems on agentic workflows, software engineering and cybersecurity, and identifies Space Bunny, an anonymous system that processed more than 200 trillion tokens, as an early version of the model.
The benchmark claims should be read as vendor-reported results, not a guarantee that every agent will complete a task reliably. MiniMax reports a 76.2 score on BankerToolBench, 73.7% on DeepSWE v1.1, 61.7 on OSWorld V2 and 29.6 on Agents’ Last Exam. The tests measure useful parts of agent performance, including planning, computer operation and software work, but production outcomes also depend on the permissions, data quality and guardrails around the model.
Speed and Cost are the Central Pitch
For many agent products, a model’s response speed matters as much as its benchmark standing. A system that must inspect documents, call tools and revise code can turn a few seconds of generation time into a frustratingly long workflow. MiniMax says M3.1 Flash can reach more than 200 output tokens per second in everyday use, with a median speed above 170 tokens per second.
Those are company figures and will vary by API load, prompt length, location and the way a developer structures an agent loop. Still, the pricing is low enough to make the comparison meaningful for teams that run high volumes of routine tasks. At the listed rate, one million output tokens cost $0.60, while cached input is priced at one tenth of the standard input rate. The model’s economics will be most appealing where teams can reuse context efficiently instead of repeatedly sending large prompts.
Weights and a Pro Model are Still Ahead
MiniMax has not released the model weights yet, saying they will arrive later. It has also previewed a more powerful M3.1 Pro model without offering a release date or specifications. That leaves developers with a hosted model today and an open-weight option still to come.
The release reflects a broader shift in the AI market toward specialized deployment choices. Frontier models may lead on some difficult tasks, but a cheaper model with fast generation can be more useful for an agent that handles support queues, document processing, code maintenance or structured back-office work at scale. MiniMax is betting that M3.1 Flash can be that practical middle ground, provided its claimed agent performance holds up outside the company’s own evaluations.
Disclaimer: AIstify is an independent media brand owned and operated by NuvexMedia LLC, publishing news, research, and insights on artificial intelligence, emerging technologies, automation, and related industries. NuvexMedia LLC invests in and collaborates with companies across the AI, technology, software, and digital innovation sectors. These relationships do not influence AIstify’s editorial coverage, and the publication maintains full editorial independence to provide accurate, timely, and objective information. © 2026 NuvexMedia LLC. All rights reserved. This content is for informational purposes only and should not be considered legal, tax, investment, financial, or other professional advice.
