MiniMax presentó H3 el 31 de julio. La combinación es poco habitual: un precio comercial de API y un plan para abrir los pesos del modelo.
H3 acepta texto, imágenes, video y audio dentro de un mismo contexto. Genera hasta 2K y 15 segundos, con audio estéreo nativo.
La apertura aún no es un archivo
“We plan to open up the model weights in the coming days, subject to applicable laws and regulations.”
Developers still need the license, commercial-use terms, hardware guidance and a technical report before treating H3 as a deployable open model.
El precio cambia la conversación
Shanghai Securities News, The Beijing News and 21st Century Business Herald reported H3 at RMB 0.8 per second for 2K generation, about one-third of comparable flagship video models. A 15-second 2K clip is roughly RMB 12 before platform packaging.
La apuesta técnica es contexto unificado
MiniMax describes H3 through Contextual Omni Representation, H3-VAE, H3-Omni Transformer and In-Context Regeneration. The model handles video, audio, images and text as one creative context.
Most source material requires about 100,000 tokens of inference before being distilled to roughly 4,000 tokens on average. H3-VAE provides a 4x gain in effective sequence length, and workload separation lifted training throughput by nearly 30%.
The next checks are whether weights appear, whether the license works for companies, and whether the community can reproduce usable text, motion and audio control on non-official hardware.
Fuentes: blog oficial de investigación de MiniMax, Shanghai Securities News, The Beijing News, CocoLoop, 21st Century Business Herald; sitio de MiniMax para verificar 2K/15 segundos, estéreo nativo, compromiso de pesos abiertos, precio de 0,8 yuanes por segundo, comparación 768p y mejora cercana al 30% en throughput de entrenamiento.