December 18, 2011 - 1AI has learned from theByteDanceIt's official, at today's Volcano Engine Force conference, ByteDance officially released theBean curdVisual Understanding Models, Multimodal Large Modeling Capabilities for Enterprises. Beanbag Visual Understanding ModelThousands of tokens input at only 3%.The company's newest product is a new, more affordable version of the original, which can handle 284 720P images for one dollar and is officially claimed to be 85% cheaper than the industry price.
The Beanbag 3D generative model was also officially unveiled at the event. Combined with veOmniverse, the Volcano Engine digital twin platform, it accomplishes the followingIntelligent training, data synthesis and digital asset productionIt is officially called "a suite of physical world simulators that support AIGC authoring".
A number of products under the Beanbag Grand model have also received updates:
- Beanbag universal model pro:Full alignment GPT-4oThe price of using the system is only 1/8th of the price of the latter;
- Music modeling: can be generated 3-minute full-length work;
- Wensheng Diagram Model version 2.1: accurately generate Chinese characters and one-sentence P-diagrams, which has been connected to Imagine AI and Doubao App.
In addition, Doubao will launch version 1.5 of its video generation model with longer video generation capability next spring, and Doubao's end-to-end real-time voice model will soon be online, unlocking new capabilities such as multi-character interpretation and dialect conversion.