##字节豆包大模型再升级，综合能力提升20.3%，火山引擎成立零售大模型生态联盟
**上海，2024年8月21日** – 2024火山引擎 AI 创新巡展在上海举办，字节跳动旗下人工智能平台火山引擎宣布豆包大模型取得重大进展，其综合能力相比三个月前首次发布时提升了20.3%。此次升级涵盖语音模型、视觉模型、对话式 AI 实时交互解决方案等多个方面，并发布了豆包·文生图模型、豆包·语音识别模型等新模型。
火山引擎总裁谭待表示，自5月15日正式对外发布以来，豆包大模型的日均 tokens 使用量已经超过5,000亿，平均企业客户使用量增长了22倍。最新版豆包大语言模型在角色扮演、语言理解、长文任务、数学、专业知识、代码能力等方面都有显著提升。
其中，豆包·文生图模型对长文本有更精准的图文匹配能力，能够创造更具美感的中国风图片。豆包·语音识别模型基于大语言模型丰富的知识和推理能力，提升了语音识别准确性，支持多种方言识别。豆包·语音合成模型升级了流式语音合成能力，能够实时响应、精准断句。
为了加速企业 AI 落地，火山引擎携手多点 DMALL 成立了零售大模型生态联盟，并宣布 AI 创造者大赛开赛。该联盟旨在通过融合豆包大模型与 AI 能力，帮助零售企业以极低的试错成本将大模型技术应用到业务场景中，推动零售行业的智能化升级。
多点 DMALL 创始人、物美集团创始人张文中博士表示，零售大模型生态联盟对于零售企业来说是抱团取暖，共享联盟内的技术成果和最佳实践，降低企业成本，是当下零售企业拥抱 AI 的最好选择。
除了零售大模型生态联盟，汽车大模型生态联盟也迎来了领克汽车、吉利银河、几何汽车等新成员，进一步壮大了生态圈。
火山引擎表示，将持续提升豆包大模型的能力，并与行业伙伴共同探索更多场景的 AI 重构，加速大模型在各行业的应用落地，推动产业智能化升级。
英语如下：
##ByteDance’s Doubao Large Language Model Upgraded, Comprehensive Capabilities Enhanced by20.3%
**Keywords:** Doubao Upgrade, AI Interaction, Volcano Engine
**Shanghai, August 21, 2024** – The 2024 Volcano Engine AI Innovation Tour was heldin Shanghai, where ByteDance’s artificial intelligence platform, Volcano Engine, announced significant progress in its Doubao large language model. The model’s comprehensive capabilitieshave improved by 20.3% compared to its initial release three months ago. This upgrade encompasses multiple aspects, including speech models, vision models, conversational AI real-time interaction solutions, and the release of new models such as Doubao Text-to-Image and Doubao Speech Recognition.
Tan Dai, President of Volcano Engine, stated that since its official launch on May 15th, Doubao’s daily token usage has exceeded 500 billion, with average enterprise customer usage increasing 22 times. The latest version of Doubao’s large language model exhibits notable improvements in areas like role-playing, language comprehension, long-text tasks, mathematics, professional knowledge, and code capabilities.
Specifically, the Doubao Text-to-Image model boasts enhancedaccuracy in matching text with images, enabling the creation of more aesthetically pleasing Chinese-style pictures. The Doubao Speech Recognition model leverages the vast knowledge and reasoning capabilities of the large language model to improve speech recognition accuracy, supporting recognition of multiple dialects. The Doubao Speech Synthesis model has upgraded its streaming speech synthesis capabilities,allowing for real-time responses and accurate punctuation.
To accelerate the implementation of AI in businesses, Volcano Engine has partnered with Duoduo DMALL to establish the Retail Large Model Ecosystem Alliance. The alliance aims to integrate Doubao’s large language model with AI capabilities, enabling retail companies to apply large model technology tobusiness scenarios with minimal trial and error, thereby driving intelligent upgrades in the retail industry.
Dr. Zhang Wenzhong, founder of Duoduo DMALL and founder of Wumart Group, expressed that the Retail Large Model Ecosystem Alliance provides a platform for retail companies to share technical achievements and best practices, reducing costs andmaking it the ideal choice for embracing AI.
Beyond the Retail Large Model Ecosystem Alliance, the Automotive Large Model Ecosystem Alliance has welcomed new members including Lynk & Co, Geely Galaxy, and Geometry, further expanding the ecosystem.
Volcano Engine has pledged to continuously enhance Doubao’s capabilities and collaborate with industrypartners to explore AI reconstruction in more scenarios, accelerating the application of large models across industries and driving intelligent upgrades in the industrial landscape.
【来源】https://mp.weixin.qq.com/s/nzNkPQqSTSA07OVytSOs7w
Views: 0