MiniMax M2
The MiniMax M2 family from MiniMaxAI introduces efficient Mixture-of-Experts models optimized for coding, agentic workflows, and tool-driven tasks. Each model features 229B total parameters with 10B activated parameters, enabling efficient execution of complex developer and automation tasks.
4 variants4 quantized variants
MiniMax M2.7
MiniMax
Params: 229B
MiniMax M2.7 is a large-scale Mixture-of-Experts (MoE) language model developed by MiniMax for coding, agentic workflows, and enterprise automation tasks. It features a 229B parameter architecture with ~10B active parameters, built on a 62-layer transformer with 48 attention heads and a 3,072 hidden size. It employs a 256-expert MoE architecture with 8 experts per token and supports up to a 200K context window. It is optimized for autonomous agent teams, iterative self-improvement, and complex real-world task execution.
Text GenerationInstruction Following+7
MiniMax M2
MiniMax
Params: 229B
MiniMax M2 is a large-scale Mixture-of-Experts (MoE) language model developed by MiniMax optimized for coding and agentic workflows. It features a 229B parameter architecture with ~10B active parameters, using 256 routed experts with 8 experts activated per token. The model is built on a 62-layer transformer with a 3072 hidden size, 48 attention heads (8 KV heads), and supports a 200K token context window. It is designed for efficient, low-latency deployment, excelling in multi-step coding tasks, tool use, and long-horizon agent execution.
Text GenerationInstruction Following+7
MiniMax M2.1
MiniMax
Params: 229B
MiniMax M2.1 is a large-scale Mixture-of-Experts (MoE) language model developed by MiniMax focused on real-world agent reliability. It features a 229B parameter architecture with ~10B active parameters, built on a 62-layer transformer with 48 attention heads and a 3,072 hidden size, utilizing 256 experts with 8 experts activated per token. It supports a 200K token context window, making it well suited for long-horizon workflows, composite instructions, and stable performance across agent frameworks and production environments.
Text GenerationInstruction Following+7
MiniMax M2.5
MiniMax
Params: 229B
MiniMax M2.5 is a large-scale Mixture-of-Experts (MoE) language model developed by MiniMax for coding, agentic workflows, and enterprise automation tasks. It features a 229B parameter architecture with ~10B active parameters, built on a 62-layer transformer with 48 attention heads and a 3,072 hidden size, utilizing 256 experts with 8 experts activated per token. The model integrates Multi-Token Prediction (MTP) for faster generation and supports a 200K token context window, enabling efficient reasoning, tool interaction, and complex task decomposition in large-scale agentic systems.
Text GenerationInstruction Following+7