Bytes beating its volcanic engine to release Doubao Model 1.5, and it is fully launched on the volcanic Ark platform. This model has achieved significant performance improvement in many fields such as knowledge, code, reasoning, and Chinese, and comprehensive capabilities have reached the world's leading level. Multiple versions surpassed the first-class models such as GPT-4O, Claude3.5sonNet in authoritative evaluations. This update not only made breakthroughs in performance, but also adhered to the principle of "increasing the price without price increase" in terms of price, and is committed to promoting the inclusiveness of AI technology.
On January 22, 2025, the volcanic engine of the byte beating announced the official release of the Doubao model 1.5 and fully launched the volcanic Ark platform. The Danbao Model 1.5 released this time has achieved significant performance improvements in multiple fields. The overall ability has reached the world's leading level, marking another important breakthrough in the artificial intelligence field in the field of artificial intelligence.
Da Bao Model 1.5 includes multiple versions. Among them, Double-1-PRO has achieved the best results in many authoritative evaluation beans such as knowledge, code, reasoning, and Chinese. First -class model. Doubao-1.5-LITE has performed well in the lightweight language model. Its performance can even be compared with the Double-PRO-32K-0828 version before the shoulder, providing users with higher cost performance. In addition, Double-1.5-VISION-PRO has been fully upgraded in multi-mode data synthesis, dynamic resolution, and multi-mode alignment, which enhances visual reasoning and fine-grained information understanding capabilities. Get leading performance.
The Danbao Model 1.5 released this time also launched the bean bag and real -time voice model, which realizes the end -to -end voice dialogue. It has the characteristics of low latency and interruption in the conversation, which brings new breakthroughs to the field of voice interaction. The volcanic engine plans to launch the corresponding API services through the Ark platform in the first half of the year, further promoting the widespread application of voice technology.
Daquan model 1.5 uses a large -scale sparse MOE architecture in the technical architecture. Through the DENSE model performance of the equivalent of 7 times the activation parameter through smaller activation parameters, it far exceeds the routine efficiency in the industry. At the same time, the self -developed server cluster scheme and network card technology have greatly reduced hardware costs, optimized the efficiency of small bag communication, and ensured the stability and efficiency of multi -machine distributed reasoning. In addition, the Dan Bao Model 1.5 did not use any other model data generated by any other model during the training, and established a complete independent data production system to ensure the independence and reliability of the source of the data.
值得注意的是，尽管豆包大模型1.5在性能和功能上实现了显着提升，但其价格保持不变，继续坚持“加量不加价”的原则，致力于推动AI技术的普惠化，让More companies and developers can benefit from this advanced technological achievement.
Experience address: https: //team.doubao.com/doubao_1_5_pro
The release of the Danbao Model 1.5 shows the powerful strength and technological innovation capabilities of byte beating in the field of artificial intelligence. Its strategy of "increased quantity and no price" also provides strong support for the popularization of artificial intelligence technology. It is believed that the Doubao model will play a greater role in the future.