MiniMax launches M3.1 Flash with lower pricing for agent workflows
The AMW Read
Published API pricing and a claimed 70% reduction in agent-task costs meaningfully update MiniMax's competitive position, though customer-level savings remain unverified.
MiniMax launches M3.1 Flash with lower pricing for agent workflows
China-based MiniMax released M3.1 Flash on October 19 through its API, M Plan subscription and MiniMax Code desktop app. API pricing is $0.15 per million input tokens, $0.60 per million output tokens and $0.015 per million cached input tokens. The company says overall agent-task usage costs are 70% below the previous model and online inference is nearly three times faster. Model weights are planned for later release, with no date disclosed.
The release positions MiniMax within the foundation-model market's competition to supply the models behind coding, computer use and business automation. For agents that repeatedly call models and tools, inference price and speed affect the economics of completing a workflow. MiniMax reports scores of 73.7% on DeepSWE v1.1, 61.7 on OSWorld V2 and 29.6 on Agents' Last Exam, alongside output throughput exceeding 170 tokens per second at the median. These are company-reported results; they support its positioning but do not establish reliable performance across customer workflows. Availability through both an API and MiniMax Code gives the release a developer integration channel and a direct coding surface.
Builders should test the claimed savings against completed tasks in their own environments. MiniMax says the model can execute workflows across multiple SaaS applications in functions including sales, customer service and finance, but the reported benchmarks do not establish each deployment's success rate or supervision needs. A concrete evaluation should compare total token spend, completion time and human intervention for the same workflow. That would show whether lower inference prices translate into lower operating costs.
