MiniMax H3 now available on AI Gateway

MiniMax H3 is now available on AI Gateway. The model generates 2K video from a text prompt, a starting image, a first-and-last frame pair, or reference material, so a single hosted endpoint covers several generation paths that previously sat in separate product surfaces.

On the product side, H3 supports text-to-video and first-frame image-to-video, plus first-to-last keyframe transitions and multimodal reference-to-video. In reference mode, one request can condition the output on reference images, video, or audio together, rather than forcing a single modality or a multi-step pipeline.

For engineers building video features, the practical win is interface consolidation. Apps that already sit on AI Gateway can call MiniMax H3 without standing up a separate MiniMax integration, and they can route text-only, image-seeded, keyframe-constrained, or multimodal-reference jobs through the same gateway path and auth model.

That placement also matters in market terms. Video generation is no longer only a chat-demo feature; it is something product teams want behind a stable API with consistent routing. Putting H3 on AI Gateway puts MiniMax’s 2K model next to whatever else teams already call through that layer, which lowers the cost of trying keyframe and reference workflows in production prototypes.

What to watch next is how far reference and keyframe modes hold up under real product constraints: whether first-to-last transitions stay coherent at 2K, whether multimodal reference (image, video, and audio in one request) actually reduces re-rolls, and whether gateway latency and quotas make H3 usable as a default path rather than a side experiment.