速递｜火山引擎Force大会：豆包大模型1.6全球创新“区间定价”，火山引擎全力打造“AI云”

Core Insights - ByteDance's subsidiary Volcano Engine held the Force Power Conference on June 11, launching new models including Doubao Model 1.6 and video generation model Seedance 1.0 Pro, while also upgrading its AI Agent development platform and cloud-native services [1][3]. Model Performance and Market Position - Doubao 1.6 has achieved top-tier global rankings in various authoritative assessments, excelling in complex reasoning, competitive mathematics, multi-turn dialogue, and instruction adherence [3]. - Doubao 1.6 supports multi-modal understanding and graphical interface operations, enabling it to handle real-world problems effectively. It can perform tasks such as automatically booking hotels and organizing shopping receipts into Excel [3][4]. - The Doubao model series has a daily token usage exceeding 16.4 trillion, a 137-fold increase since its initial release in May last year. It holds a 46.4% market share in China's public cloud large model market [4]. Technological Innovations - Doubao 1.6 integrates multi-modal capabilities, visual understanding, deep thinking, and basic language models into a single enhanced version, focusing on pure thinking scenarios [5]. - The introduction of "input length interval pricing" significantly reduces costs, with the base model priced at 0.8 yuan per million tokens for input and 8 yuan per million tokens for output, making it one-third the cost of Doubao 1.5 or DeepSeek-R1 [5][6]. Business Strategy and Ecosystem Development - Volcano Engine has launched a full-stack AI cloud-native service, including products like MCP service, PromptPilot, and veRL framework, aimed at lowering the development threshold for AI agents [7]. - The company emphasizes that over 80% of its engineers are using AI programming tools like TRAE, which has over 1 million monthly active users [9]. Future Outlook - The AI era is expected to transform development paradigms and technical architectures, with agents becoming proactive executors capable of complex task completion [10][11].