MiniMax launches M3.1 Flash with lower-cost inference
MiniMax has released M3.1 Flash, promising three-times faster responses and a 70% price cut as competition shifts toward efficient coding models.
What happened
MiniMax released M3.1 Flash on October 19, positioning it as a faster and cheaper member of its M-series for software development and general agent workloads. The company says the model is three times faster than its predecessor and reduces pricing by 70%. MiniMax also linked the launch to Space Bunny, an earlier model version that the company says processed more than 200 trillion tokens and reached leading positions on global benchmarks.
The timing matters because the release arrives during an unusually intense cycle for low-cost coding models. Rather than presenting a new frontier-scale model, MiniMax is competing on the operating variables that determine whether developers use a model repeatedly: latency, token economics and practical performance in long coding sessions. Artificial Analysis Arena data cited in the launch coverage placed M3.1 Flash at number 12 among open models and number 32 overall in its WebDev evaluation, with a score of 1,588 points.
Why it matters
The release also reveals a two-stage strategy. MiniMax is making the model available through its service and coding products first, while public discussion around the launch indicates that open weights are expected later, on October 23. That means today’s event is a service launch, not yet an open-weight release. The distinction is important for developers deciding whether they can self-host the model or must rely on MiniMax’s infrastructure.
For the market, the price cut is more consequential than the branding. If the reported speed and coding results hold under independent testing, M3.1 Flash could pressure other Asian model providers to lower inference prices or improve throughput. The unresolved question is whether its advantage survives demanding multi-file tasks, tool use and long-context workloads outside benchmark-style WebDev tests.