MiniMax: MiniMax M2.5

MiniMax-M2.5 is a state-of-the-art large language model engineered for unparalleled real-world productivity. Building on the coding prowess of M2.1, this model extends its capabilities to general office work, demonstrating fluency in generating and operating Word, Excel, and PowerPoint files. It excels at context switching between diverse software environments and collaborating across different agent and human teams. Achieving impressive scores of 80.2% on SWE-Bench Verified, 51.3% on Multi-SWE-Bench, and 76.3% on BrowseComp, M2.5 is also remarkably token efficient. Its training optimizes actions and output through sophisticated planning. With a generous context window of 196K tokens and a max output of 65K tokens, MiniMax M2.5 offers robust performance for complex tasks. Pricing is set at $0.30 per 1M input tokens and $1.10 per 1M output tokens, available on the PRO access tier. Key capabilities include json_mode, streaming, functions, long_context, and structured output, making it highly versatile for various applications requiring precise and efficient AI interaction.