Chinese AI startup MiniMax has launched an updated model called M3.1 Flash. The company says it will bring major changes in coding, cybersecurity, and tasks that need AI tools to be done in multiple steps. MiniMax also claims that the model will respond nearly three times faster than earlier versions, along with the cost of using it falling by 70%. The launch adds to the race among AI firms to build tools that work faster and cost less.
MiniMax says M3.1 Flash can help write code, fix software issues, use computer tools, and handle tasks across work apps. The company has also reported strong results in its own tests. However, these claims still need to be checked through independent tests and everyday use. The model is available through MiniMax’s API, M Plan, and MiniMax Code. MiniMax says it plans to release the model files later, so developers can access them beyond its current services.
Writing code can take time, especially when developers need to find bugs or fix errors. A faster AI model could help them finish these tasks sooner. It may also handle more steps without needing constant input from a person. This could help teams spend less time on routine work. Still, speed is only one part of the picture. The model must also give correct answers and produce code that works.
Also Read: Microsoft Brings AI Agents to PCs with New Coding, Security Tools
Lower AI Costs Could Help Startups and Big Firms
AI models use computing power each time they read a request and create an answer. The cost of doing this is called an inference cost. MiniMax lists M3.1 Flash at USD 0.15 per million input tokens and USD 0.60 per million output tokens. Tokens are small pieces of text that an AI model reads or creates.
For one million input tokens and one million output tokens, the listed price adds up to USD 0.75. At the same rates, MiniMax’s M3 model would cost about USD 1.50 for that workload. This example shows how the lower price could help teams that send large numbers of requests.
Startups will be most beneficial if the model really delivers what it promises, though they have to check the model quality and speed before implementing it completely. M3.1 Flash makes a strong claim on price and performance, but real-world use will show how much it can deliver.