MiniMax M2.7模型登陆NVIDIA GPU加速端点，支持NemoClaw与OpenClaw

MeNews · 2026-04-12T07:20:48+00:00

MiniMax announces that its M2.7 model is now available on NVIDIA GPU-accelerated endpoints, supporting multiple tools. M2.7 is a sparse mixture-of-experts model optimized for agent workflows and complex applications, with 230 billion parameters and a 200k-token context length. NVIDIA has released an open-source reference stack to support secure deployment.

MeNews

2026-04-12 07:20:48

Abstract generation in progress

ME News Message, April 12 (UTC+8), MiniMax announced that its M2.7 model is now available on NVIDIA GPU-accelerated endpoints and supports tools including NemoClaw and OpenClaw. M2.7 is the latest version in the M2 series, designed to enhance agent workflows and inference, as well as complex applications such as machine learning research. It is a sparse mixture-of-experts model with a total parameter count of 2300亿, an active parameter count of 100亿 per token, and an activation rate of approximately 4.3%. It supports a context length of 20万 tokens, has 256 local experts, activates 8 experts per token, and has 62 layers. To support deployment, NVIDIA launched the open-source reference stack NemoClaw to simplify the secure deployment of OpenClaw, and it can be started with one click via the NVIDIA Brev cloud platform. In addition, NVIDIA has also collaborated with the open-source community to integrate high-performance kernels optimized for large-scale MoE models in frameworks such as vLLM and SGLang. (Source: InFoQ)

This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.