Introducing MiniMax M2.5
We are pleased to announce that ChatBotKit now supports MiniMax M2.5, the productivity-focused flagship model from MiniMax, now available across all ChatBotKit platform features including bots, blueprints, skillsets, and the API.
MiniMax M2.5 is engineered with a clear focus on practical, high-volume workloads. Rather than optimizing purely for benchmark headlines, it targets the real-world tasks that teams rely on every day - coding assistance, office automation, document generation, and multi-step agent execution. The result is a model that performs strongly across these categories while maintaining high token efficiency, making it a compelling choice for applications that need reliable performance at scale.
One of the most notable characteristics of MiniMax M2.5 is its pricing. With an input token cost of $0.30 per million tokens and output at $1.20 per million, it delivers competitive benchmark performance at a fraction of what comparable models cost. For teams running high-throughput workflows - automated pipelines, large-scale content generation, or frequent agent interactions - this value proposition translates directly into meaningful cost savings without sacrificing quality.
MiniMax M2.5 supports a 204,800-token context window, providing ample room for multi-turn conversations, long document analysis, and extended code generation sessions. Its strong performance in multi-step agent workflows makes it well-suited for ChatBotKit Blueprint-based architectures where an agent must plan, execute, and adapt across numerous sequential operations.
Function calling support allows MiniMax M2.5 to interact seamlessly with external tools and APIs through ChatBotKit Skillsets. Whether you are building a document processing assistant, a customer support automation, or a complex multi-tool research agent, MiniMax M2.5 provides the reasoning depth and execution reliability to handle the workflow end-to-end.
To get started with MiniMax M2.5, select it from the model picker in your ChatBotKit dashboard or reference it via the API.