MiniMax stellte H3 am 31. Juli vor. Ungewöhnlich ist die Kombination aus kommerziellem API-Preis und angekündigter Öffnung der Modellgewichte.
H3 verarbeitet Text, Bilder, Video und Audio in einem Kontext. Es erzeugt bis zu 2K Auflösung und 15 Sekunden Länge mit nativem Stereo.
Offene Gewichte sind noch kein Download
“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.
Der Preis macht den Angriff greifbar
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.
Die technische Wette heißt einheitlicher Kontext
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.
Quellen: offizieller MiniMax-Forschungsblog, Shanghai Securities News, The Beijing News, CocoLoop, 21st Century Business Herald; MiniMax-Website zur Prüfung von 2K/15 Sekunden, nativem Stereo, Open-Weight-Zusage, RMB 0,8 pro Sekunde, 768p-Vergleich und fast 30% mehr Trainingsdurchsatz.