广告
加载中

代码和视频 火山引擎一个没落下

胡镤心 2026-06-23 15:14
胡镤心 2026/06/23 15:14

邦小白快读

EN
全文速览

本次火山引擎2026夏季原动力大会发布了多款面向生产级应用的AI新产品,披露了大模型市场的最新增长数据,普通用户可以从中获得以下实用干货:

1.核心产品性能突出,国产大模型已经达到国际一线水平,本次发布的豆包2.1 Pro在编程、智能体、视觉语言三大领域跨越生产级质变点,多项评测成绩优于国际竞品Claude Opus 4.6,已经开放API服务,接入多个主流产品可直接使用;

2.价格优势非常明显,豆包2.1 Pro综合使用成本比国际竞品低近80%,面向高频场景的豆包2.1 Turbo价格再降一半,普通开发者和个人都能用得起;

3.还有多款内容创作模型即将上线,7月上线的Seedance 2.5视频生成模型可直接输出30秒4K高清视频,支持多素材生成和局部编辑,搭配图像、音频生成模型,可满足各类内容创作需求。

本次大会披露了大模型产业的最新发展趋势,能帮助品牌把握AI升级方向,落地AI应用降本增效,核心干货如下:

1.行业消费与产业趋势明确,当前大模型已经从技术比拼进入规模化商用阶段,市场需求已经从娱乐聊天转向生产级交付,大模型从消费端逐步走向产业端,全行业对AI的需求高速增长,日均Token调用量一年增长超10倍,200家年调用超万亿的企业覆盖互联网、制造、金融、汽车等多个行业,AI融入核心业务的速度正在加快;

2.可落地的AI能力可支撑品牌多场景需求,大模型的生产级代码能力可帮助品牌快速开发数字化业务系统,多模态生成能力可支撑品牌低成本快速产出营销图文、视频内容,Seedance 2.5就能直接输出符合营销需求的高清视频,大幅降低内容生产门槛;

3.国产大模型成本优势明显,综合使用成本比国际一线模型低80%,能大幅降低品牌布局AI的资金压力。

本次大会披露了大模型MaaS市场的最新变化,给AI服务相关卖家指明了新的增长机会,核心干货整理如下:

1.市场增长空间巨大,当前中国公有云MaaS市场已经进入爆发期,火山引擎以49.5%的份额位居市场第一,全行业需求高速增长,豆包大模型日均Token调用量过去一年增长超10倍,年Token调用超1万亿的企业数量半年翻倍,达到200家,AI融入企业核心业务的速度还会继续加快,围绕AI生产服务的相关卖家会获得大量增长机会;

2.明确的业务落地方向,卖家可以围绕生产级大模型的能力,开展代码开发、智能体搭建、内容生成等相关业务,火山引擎已经开放全系列模型的API服务,还推出了配套的开发工具,卖家可以直接接入,无需高额自研成本;

3.成本优势可帮助卖家提升竞争力,大模型本身价格远低于国际竞品,卖家可以更低的成本为客户提供服务,获得价格优势,同时还有不同价位的产品覆盖高低频不同场景,适配不同客户的需求。

本次大会发布的生产级大模型能力,给工厂推进数字化转型、落地AI应用提供了新的支撑,核心干货如下:

1.可直接落地生产场景的AI能力已经成熟,当前大模型已经进入生产级应用阶段,视频生成模型Seedance已经落地工业制造领域,可以为工厂提供数据合成、场景仿真、生产流程演示的工具能力,帮助工厂降低研发测试环节的成本,提升研发效率;豆包2.1 Pro的生产级代码能力,已经完成了芯片设计RTL测试的完整工程流程,可支撑工厂的工业软件开发、复杂产品研发等任务;

2.给工厂带来了新的商业机会,当前已经有大量制造企业接入大模型服务,大模型技术的成熟让工厂可以打造全新的智能化生产模式,开发新的增值服务,拓展营收空间;

3.给工厂推进数字化带来新启示,工厂不需要投入高额成本自研大模型,可以借助成熟的MaaS服务快速落地AI应用,降低转型门槛和风险,加快数字化升级的速度。

本次大会披露了大模型行业的最新发展趋势和技术进展,能帮助AI服务服务商把握方向、解决客户痛点,核心干货整理如下:

1.行业发展趋势清晰,当前大模型行业已经从技术探索转向规模化商用,客户的核心需求已经从“能聊天”转向“能交付”,越来越多的产业客户需要生产级的AI能力,需求规模高速增长,市场空间持续扩大,服务商锚定生产级AI服务就能抓住行业红利;

2.掌握了最新的技术成果可直接使用,国产豆包2.1 Pro已经在编程、智能体、多模态领域跨越生产级质变点,核心性能超过国际一线竞品,综合成本低80%,同时已经形成覆盖代码、视频、图像、音频的全品类多模态大模型矩阵,能满足不同行业客户的各类需求;

3.有成熟的配套解决方案可复用,火山引擎已经推出一整套生产级AI方案,包含升级后的AI云原生架构、Agent开发工具、企业级Agent工作台、安全产品体系,服务商可以基于这套方案快速为客户打造定制化AI解决方案,大幅降低自身的开发成本,提升交付效率。

本次大会披露了企业级AI服务市场的需求变化,以及头部平台的最新运营做法,给大模型平台的发展提供了参考,核心干货如下:

1.当前市场需求已经发生明确变化,企业客户对AI的需求已经从尝鲜式的聊天应用升级为生产级业务交付,对模型的稳定性、长程任务处理能力、多模态理解能力要求更高,同时对成本控制要求高,市场整体需求高速增长,目前已经有超过110万企业和个人使用大模型服务,需求还在快速扩张;

2.头部平台的成熟做法可参考,火山引擎通过搭建覆盖多场景的全品类大模型矩阵,持续快速迭代模型能力,匹配不断升级的客户需求,同时配套推出低门槛的开发工具,降低开发者和企业的接入难度,再通过有竞争力的定价抢占市场,目前已经拿下近半数的中国公有云MaaS市场份额;

3.需要规避发展风向,当前大模型技术迭代速度快,客户需求升级快,平台如果不能持续跟进技术升级、匹配生产级需求,很容易被市场淘汰,需要持续投入技术迭代,跟上市场变化。

本次大会披露了中国大模型产业的最新发展数据和进展,对研究大模型产业发展有很高的参考价值,核心干货整理如下:

1.产业发展出现明确新动向,中国大模型产业已经完成了从技术验证到规模化商用的转型,产业需求结构发生本质变化,从原来消费端的聊天娱乐需求为主,转向产业端的生产级交付需求为主,国产大模型已经在核心能力上追上国际一线水平,能够和国际头部模型正面竞争;

2.产业发展呈现新特征,MaaS模式已经得到市场的充分验证,目前火山引擎以49.5%的市场份额位居中国公有云MaaS市场第一,行业规模增长速度惊人,豆包大模型日均Token调用量两年涨幅达到1500倍,过去一年增长超10倍,年调用量超万亿的企业数量半年翻倍,说明产业端需求已经进入爆发阶段;

3.商业模式已经跑通,按调用量收费的MaaS模式已经被市场接受,差异化定价策略可以覆盖不同场景、不同规模客户的需求,低价策略进一步加速了AI在产业端的普及,为产业研究提供了新的样本。

返回默认

声明:快读内容全程由AI生成,请注意甄别信息。如您发现问题,请发送邮件至 run@ebrun.com 。

我是 品牌商 卖家 工厂 服务商 平台商 研究者 帮我再读一遍。

Quick Summary

At the 2026 Summer Yuan Dongli (Original Power) Conference hosted by Volcano Engine, the company launched a number of new production-grade AI products and released the latest growth data for China's large language model (LLM) market. For general users, the key takeaways are:

1. Domestic Chinese LLMs have now reached the top international level in core performance. The newly launched Doubao 2.1 Pro crossed the production-grade threshold in three key domains: coding, intelligent agents, and vision-language processing. It outperforms the international competitor Claude Opus 4.6 on multiple benchmarks, and its API service is already open for integration into most mainstream products for immediate use.

2. It offers a dramatic price advantage: the total cost of using Doubao 2.1 Pro is nearly 80% lower than that of its international competitors. For high-frequency use cases, Doubao 2.1 Turbo cuts prices by an additional 50%, making it affordable for ordinary developers and individual users.

3. A range of new content creation models will launch soon. Seedance 2.5, a video generation model launching in July, can directly output 30-second 4K HD videos, supports multi-source generation and local editing. When paired with Volcano Engine's image and audio generation models, it can meet all types of content creation needs.

The conference outlined the latest development trends of the large model industry, helping brands map out their AI upgrade paths and implement AI applications to cut costs and boost efficiency. Key takeaways for brands include:

1. Clear industry and consumer trends have emerged: LLMs have moved past pure technology competition and entered the stage of large-scale commercialization. Market demand has shifted from casual chatbot use cases to production-grade delivery, with large models gradually expanding from consumer-facing use cases to enterprise and industrial scenarios. Industry-wide demand for AI is growing rapidly: daily token consumption has grown more than 10-fold in one year, and 200 enterprises across the internet, manufacturing, finance, automotive and other sectors now exceed 1 trillion tokens in annual usage. The integration of AI into core business operations is accelerating.

2. Production-grade AI capabilities can support brands' multi-scenario needs. LLM's production-grade coding capabilities help brands quickly build digital business systems, while multimodal generation capabilities allow brands to produce marketing copy, images and videos quickly at low cost. Seedance 2.5 can directly output HD videos that meet marketing requirements, drastically lowering the barrier to content production.

3. Domestic large models offer significant cost advantages, with total usage costs 80% lower than top international models, greatly reducing the financial pressure for brands to adopt AI.

The conference revealed the latest changes in the LLM-as-a-Service (MaaS) market, pointing to new growth opportunities for AI service sellers. Key takeaways are as follows:

1. The market offers enormous growth potential. China's public cloud MaaS market has entered a period of explosive growth, with Volcano Engine ranking first with a 49.5% market share. Industry-wide demand is growing rapidly: daily token usage of Doubao LLM grew more than 10-fold over the past year, and the number of enterprises with annual token usage exceeding 1 trillion doubled in six months to 200. As AI integration into enterprise core business continues to accelerate, sellers focused on AI-enabled production services will see massive growth opportunities.

2. Clear implementation paths are available. Sellers can build businesses around production-grade LLM capabilities, including code development, agent building, content generation, and other related services. Volcano Engine has opened APIs for its full model line and launched supporting development tools, allowing sellers to integrate directly without incurring high in-house R&D costs.

3. The cost advantage of these models helps sellers boost their competitiveness. Since LLMs themselves are priced far lower than international competitors, sellers can serve customers at lower cost to gain a price advantage. In addition, tiered pricing across different product lines covers both high- and low-frequency use cases to meet the needs of different customers.

The production-grade LLM capabilities launched at the conference provide new support for factories advancing digital transformation and implementing AI applications. Key takeaways are as follows:

1. AI capabilities ready for direct deployment in production scenarios are now mature. LLMs have entered the production-grade application phase. The Seedance video generation model is already deployed in industrial manufacturing, providing factories with tools for data synthesis, scenario simulation, and production process demonstration, helping cut R&D and testing costs and improve R&D efficiency. Doubao 2.1 Pro's production-grade coding capability has already completed the full engineering workflow of RTL testing for chip design, enabling it to support industrial software development, complex product R&D and other tasks for factories.

2. The technology opens new business opportunities for factories. A large number of manufacturing enterprises have already integrated LLM services. The maturation of LLM technology allows factories to build brand-new intelligent production models, develop new value-added services, and expand revenue streams.

3. It offers new insights for factories advancing digital transformation. Factories do not need to invest heavily to develop their own LLMs in-house; they can leverage mature MaaS services to implement AI applications quickly, lowering the barrier and risk of transformation and accelerating the pace of digital upgrade.

The conference revealed the latest development trends and technical progress of the large model industry, helping AI service providers align their strategies and solve customer pain points. Key takeaways are as follows:

1. The industry development trajectory is clear: the large model industry has shifted from technology exploration to large-scale commercialization. Customers' core demand has shifted from "can chat" to "can deliver production-grade results", with a growing number of enterprise clients seeking production-grade AI capabilities. Demand is growing rapidly and the market continues to expand, so providers that focus on production-grade AI services will capture industry growth dividends.

2. The latest cutting-edge technical results are available for direct integration. The domestic model Doubao 2.1 Pro has crossed the production-grade threshold in coding, intelligent agents, and multimodal processing, with core performance outperforming top international competitors at 80% lower total cost. Volcano Engine has also built out a full-stack multimodal model lineup covering code, video, image, and audio generation, capable of meeting the diverse needs of enterprise clients across industries.

3. Mature turnkey solutions are available for reuse. Volcano Engine has launched a full suite of production-grade AI solutions, including an upgraded AI-native cloud architecture, Agent development tools, an enterprise-grade Agent workbench, and an enterprise security product system. Service providers can use this framework to quickly build customized AI solutions for clients, drastically cutting their own development costs and improving delivery efficiency.

The conference revealed shifting demand in the enterprise AI service market and the latest operational strategies of leading platforms, providing valuable reference for LLM platform development. Key takeaways are as follows:

1. Market demand has changed clearly: enterprise demand for AI has upgraded from experimental chatbot use cases to production-grade business delivery. Clients now have higher requirements for model stability, long-horizon task processing, and multimodal understanding, alongside strict demands for cost control. Overall market demand is growing rapidly, with more than 1.1 million enterprises and individual users already using LLM services, and demand continues to expand quickly.

2. Mature practices from leading platforms offer clear reference. Volcano Engine built a full-stack LLM portfolio covering multiple scenarios, continuously and rapidly iterates model capabilities to match evolving customer demand, and launched low-barrier development tools to reduce integration friction for developers and enterprises, then captured market share with competitive pricing. It now holds nearly half of China's public cloud MaaS market.

3. Platforms need to avoid common pitfalls. Given the rapid pace of LLM technology iteration and fast-upgrading customer demand, platforms that fail to keep up with technical upgrades and meet production-grade demand will likely be pushed out of the market. Continuous investment in technical iteration is required to keep pace with market changes.

The conference released the latest development data and progress of China's large model industry, offering high reference value for research on LLM industry development. Key insights are as follows:

1. Clear new industry trends have emerged: China's large model industry has completed its transition from technical validation to large-scale commercialization, and the industry's demand structure has undergone a fundamental shift. Where consumer-facing chat and entertainment use cases dominated before, enterprise production-grade delivery demand now makes up the majority of the market. Domestic large models have caught up to top international models in core capabilities, and are now able to compete head-to-head with leading global models.

2. The industry shows new development characteristics. The MaaS model has now been fully validated by the market. Volcano Engine ranks first in China's public cloud MaaS market with a 49.5% share, and industry growth has been extraordinary. Daily token usage of Doubao LLM has grown 1,500x over two years, and more than 10x in the past year alone; the number of enterprises with annual token usage exceeding 1 trillion doubled in six months, confirming that enterprise demand has entered an explosive growth phase.

3. The business model is now proven. The pay-per-token MaaS model has gained broad market acceptance, and differentiated pricing can meet the needs of customers across different scenarios and scales. Low-cost pricing has further accelerated the adoption of AI in enterprise scenarios, providing a valuable new case study for industry research.

Disclaimer: The "Quick Summary" content is entirely generated by AI. Please exercise discretion when interpreting the information. For issues or corrections, please email run@ebrun.com .

I am a Brand Seller Factory Service Provider Marketplace Seller Researcher Read it again.

【亿邦原创】6月23日,火山引擎在北京举办2026夏季FORCE原动力大会。大会披露的数据显示,截至今年6月,豆包大模型日均Token调用量已突破180万亿,过去一年增长超10倍;在中国公有云MaaS服务市场,火山引擎以49.5%的市场份额位居第一。年Token调用量超过1万亿的企业已达200家,半年内翻倍,覆盖互联网、制造、金融、汽车等行业。规模增长的同时,行业对模型能力的要求也在发生变化——从能聊天转向能交付。

本次大会发布的主力模型豆包2.1 Pro,在Coding、Agent、VLM三个方向上实现能力跃升,多项评测表现优于Claude Opus 4.6。火山引擎总裁谭待认为,这是豆包首次在编程与智能体领域跨越“生产级质变点”。与此同时,视频生成模型Seedance 2.5亮相,预计7月上线,已在具身智能、工业制造等实体产业中落地。从模型能力升级到云服务基础设施迭代,火山引擎在此次大会上给出了面向生产级AI的一整套更新方案。

1、豆包2.1 Pro跨越生产级“质变点”

谭待在会上表示,只有当模型能力跨越“质变点”,才能真正满足企业与个人在生产场景中的使用需求。全球范围内,第一个跨越质变点的视频生成模型是Seedance 2.0,Coding与Agent领域则是Claude Opus 4.6。

作为本次大会发布的主力模型,豆包2.1 Pro在Coding(编程)、Agent(智能体)、VLM(视觉语言模型)三大核心方向实现能力跃升,多项评测表现优于Claude Opus 4.6,正式跨越生产级质变点。评测成绩显示,豆包2.1 Pro在Terminal Bench 2.1、SWE-Pro、SciCode等代码评测中进入国际第一梯队;在Agent与多模态方面,该模型在OSWorld、MobileWorld、MMMU-Pro等评测中也位居全球前列。

谭待以两个实践案例展示了模型的Coding和Agent能力:在一项芯片设计RTL的测试中,豆包2.1 Pro连续运行近18小时,经历9轮迭代,跑通了仿真、测试、综合检查等完整工程流程,展现出模型在真实工程场景里的生产级Coding交付能力;依托豆包2.1 Pro模型搭建3D虚拟城市场景,可实现500余个智能Agent同步协作,完成上千轮工具调用,生成超百栋建筑。

价格方面,豆包2.1 Pro每百万Tokens输入价格为6元、输出价格为30元,缓存命中价格仅1.2元。火山引擎表示,其综合使用成本较Claude Opus 4.6降低近80%。面向高频调用场景的豆包2.1 Turbo,价格进一步降至2.1 Pro的一半。同时,火山引擎上线Doubao-Seed-Evolving版本,面向Coding与Agent场景,将以每月2至4次的频率快速迭代。目前,豆包2.1模型已在火山引擎开放API服务,并陆续接入豆包、TRAE、扣子等产品。

2、Seedance 2.5亮相,多模态矩阵持续扩展

本次大会上,最新的豆包视频生成模型Seedance 2.5首次亮相,预计将于7月正式上线。据介绍,Seedance 2.5实现了30秒单段原生视频直出、最多支持50个全模态素材联合生成、保持画面一致性的局部编辑三项能力提升。与上一代Seedance 2.0最长15秒的生成时长相比,新版本直接将单次成片时长翻倍;参考素材上限也从12份大幅提升至50份。新版本还原生支持4K高清成片输出。

谭待认为,视频生成是通往世界模型的路径之一,在实体产业中有巨大的应用潜力。目前,Seedance已在具身智能、工业制造、智能驾驶等领域落地,为数据合成、场景仿真、流程演示等业务需求提供新的工具能力。

此外,火山引擎还展示了豆包图像创作模型Seedream 5.0 Pro和豆包音频生成模型Seed-Audio 1.0。Seedream 5.0 Pro具备交互式精准编辑、多图层分离、高密度信息表达和原生多语种文字生成等能力,可直接生成可编辑分层设计图和复杂信息图。Seed-Audio 1.0则具备零样本多模态参考能力,支持一次生成多角色对白、背景音乐和拟音特效。

面向Agent时代,火山引擎再次升级AI云原生架构,发布方舟CLI命令行工具,支持开发者通过一行指令将Agent接入火山方舟;同时升级AgentKit与HiAgent 3.0,并推出ArkClaw企业版Agent工作台和AI Trust产品体系。

3、火山引擎近况:Token调用量与市场份额双增长

据火山引擎披露,截至今年6月,豆包大模型日均Token调用量已突破180万亿,过去一年增长超10倍。与2024年5月1200亿的初始规模相比,两年间涨幅达1500倍。

IDC数据显示,在中国公有云MaaS服务市场,火山引擎以49.5%的市场份额位居第一。火山引擎年度累计Token消耗超万亿的“万亿Token俱乐部”成员已从2025年12月的100家增至200家。

大会披露,目前已有超过110万企业和个人使用火山方舟大模型服务。年Token调用量超过1万亿的企业已达200家,半年内增长一倍,覆盖互联网、制造、金融、汽车等多个行业。

豆包2.1 Pro在Coding与Agent方向跨过质变点,标志着国产大模型在代码交付、长程任务执行、多模态理解和企业级稳定性上,已经能和国际一线模型正面竞争。芯片设计RTL测试里连续跑了近18小时,500多个智能Agent协同搭建3D城市——这些案例说明,模型不再只是聊天工具,开始真正进入生产环境承担复杂任务。

另一方面,视频生成模型Seedance 2.5已经进了工厂和自动驾驶场景,大模型技术正在从消费端往产业端走。火山引擎目前49.5%的MaaS市场份额、180万亿的日均Token调用量也显示,中国大模型市场已经从技术比拼进入规模化商用的阶段。

接下来,随着模型能力继续提升、价格持续走低,AI融入企业核心业务的速度只会更快。

文章来源:亿邦动力

广告
微信
朋友圈

这么好看,分享一下?

朋友圈 分享

APP内打开

+1
+1
微信好友 朋友圈 新浪微博 QQ空间
关闭
收藏成功
发送
/140 0