Compare MiniMax M2.7 and MiniMax M3 on key metrics including price, context length, throughput, and other model features.
MiniMax-M2.7 is a next-generation large language model built for autonomous, real-world productivity and continuous improvement. Designed to take an active role in its own development, M2.7 incorporates advanced agent capabilities through multi-agent collaboration, allowing it to plan, execute, and improve complex tasks across dynamic environments. Trained for production-level performance, M2.7 supports workflows such as live debugging, root cause analysis, financial modeling, and full document creation across Word, Excel, and PowerPoint. It delivers strong benchmark results, including 56.2% on SWE-Pro and 57.0% on Terminal Bench 2, while reaching 1495 ELO on GDPval-AA, setting a new benchmark for multi-agent systems in real-world digital workflows.
MiniMax-M3 is a multimodal foundation model from MiniMax. It supports text, image, and video inputs with text output, a 1M-token context window, and is suited for long-horizon agentic work, coding, and tool use. It is built on MiniMax Sparse Attention (MSA), which replaces full attention with KV-block selection to cut per-token compute at long context — roughly 1/20 the cost of the previous generation at 1M tokens, with substantially faster prefill and decode while retaining quality across most tasks. Trained as a native multimodal model on interleaved data and tuned for multi-turn, production-like collaboration via an interactive user-simulator framework, the model is oriented toward sustained, multi-step tasks rather than single-turn execution.