Core Insights - Volcano Engine has released a series of updates for the Doubao large model, including Doubao model 1.6, which natively supports multiple thinking lengths, and new models such as Doubao model 1.6 lite, Doubao voice synthesis model 2.0, and Doubao voice replication model 2.0 [1] Summary by Categories - **Product Updates** - Doubao model 1.6 introduces native support for various thinking lengths [1] - New models launched include Doubao model 1.6 lite, Doubao voice synthesis model 2.0, and Doubao voice replication model 2.0 [1] - **Performance Metrics** - As of September 30, 2025, the daily average token usage for the Doubao large model exceeds 30 trillion, representing an increase of over 80% compared to the end of May [1]

Core Insights - Volcano Engine has released a series of updates for the Doubao large model, including Doubao 1.6, which natively supports multiple thinking lengths, and introduced Doubao 1.6 lite, Doubao Speech Synthesis Model 2.0, and Doubao Voice Replication Model 2.0 [1][2] Model Updates - Doubao 1.6 is the first large model in China to support "tiered adjustment of thinking length," offering four options: Minimal, Low, Medium, and High, which balance model performance, latency, and cost [3] - The upgraded Doubao 1.6 model shows a 77.5% reduction in total output tokens and an 84.6% decrease in thinking time at low thinking length, while maintaining model effectiveness [3] - Doubao 1.6 lite is lighter and faster than the flagship version, outperforming Doubao 1.5 pro by 14% in enterprise-level assessments and reducing overall usage costs by 53.3% in the most commonly used input range of 0-32k [3] Speech Models - The newly released Doubao Speech Synthesis Model 2.0 and Doubao Voice Replication Model 2.0 feature enhanced emotional expressiveness and precise instruction adherence, capable of accurately reading complex formulas [8] - These models have achieved a 90% accuracy rate in reading complex formulas for subjects from elementary to high school, addressing a significant challenge in the industry [8] Intelligent Model Routing - Volcano Engine has introduced the Smart Model Router, the first intelligent model selection solution in China, allowing users to choose from "Balanced Mode," "Effect Priority Mode," and "Cost Priority Mode" for optimal model selection based on task requests [10] - In tests, the Smart Model Router improved the effectiveness of the DeepSeek model by 14% in Effect Priority Mode and reduced overall costs by over 70% in Cost Priority Mode while achieving similar results [10] Market Position - As of September 2025, the daily token usage of Doubao large model has exceeded 30 trillion, representing an over 80% increase since May 2023 [1] - Volcano Engine holds a 49.2% market share in China's public cloud large model service market, ranking first according to IDC [1]