MiniMax-H3

MiniMax-H3-Highspeed is a faster, lower-cost tier of the omni-modal MiniMax-H3 video generation model, balancing quality and speed for everyday content creation. It produces 4-15 second clips at 480P or 768P (no 2K) with native synchronized audio. It accepts a text prompt plus optional first-frame and last-frame images, and up to 9 reference images, 3 reference videos, and 3 reference audio clips, and supports text-to-video, image-to-video, first-and-last-frame, and reference generation.

MiniMax-H3-Lite is the lightweight, lowest-cost tier of the omni-modal MiniMax-H3 video generation model, built for creative exploration and batch generation. It produces 4-15 second clips at 480P or 768P (no 2K) with native synchronized audio. It accepts a text prompt plus optional first-frame and last-frame images, and up to 9 reference images, 3 reference videos, and 3 reference audio clips, and supports text-to-video, image-to-video, first-and-last-frame, and reference generation. It does not yet support the adaptive aspect ratio, so a concrete ratio must be specified.

MiniMaxAI/MiniMax-M2.5 is a 229B-parameter Mixture-of-Experts (MoE) language model with 10B active parameters per token, designed as the world's first production-level model natively optimized for agentic scenarios like coding, tool use, search, and office productivity.