ByteDance has officially launched its latest Doubao large model 1.5Pro, which demonstrates outstanding comprehensive capabilities across multiple fields, successfully surpassing industry-renowned models like GPT-4o and Claude3.5Sonnet. The release of this model marks an important step forward for ByteDance in the field of artificial intelligence.
The Doubao 1.5Pro utilizes a new sparse MoE (Mixture of Experts) architecture, pre-trained with fewer activation parameters. The innovation of this design lies in its ability to deliver performance equivalent to a Dense model with 7 times the activation parameters, significantly exceeding the efficiency of conventional MoE architectures in the industry, achieving approximately 3 times the efficiency improvement. This design allows the Doubao large model to score exceptionally well across various assessment benchmarks in knowledge, coding, reasoning, and Chinese language processing.
In addition to the upgrades of the main model, ByteDance also released the Doubao visual understanding model Doubao-1.5-vision-pro and the Doubao real-time voice model Doubao-1.5-realtime-voice-pro. The new visual understanding model has undergone comprehensive technical upgrades in multimodal data processing, dynamic resolution, and fine-grained information understanding, further enhancing its capabilities in visual reasoning and text comprehension. Meanwhile, the launch of the real-time voice model allows the Doubao App to provide a smoother voice conversation experience, featuring low latency and the ability to interrupt conversations at any time.
ByteDance officially stated that the Doubao large model was trained without using any data generated by external models, ensuring the model's independence and reliability. Furthermore, the pricing of all new products will remain unchanged, allowing users to directly experience the new features within the Doubao App.
This release not only showcases ByteDance's ongoing innovation capabilities in the AI field but also provides developers with robust API support, further promoting the popularization and application of artificial intelligence technology.
