目录
通义千问上线“长记忆”功能，实现个性化连续对话
Tongyi Qianwen Launches "Long-Term Memory" Feature for Personalized Continuous Conversations 🧠
Wenxin Assistant Upgraded to Support 8 AIGC Modalities, Can Generate 3-Minute Videos from a Single Sentence 🚀
Doubao Large Model Daily Invocation Volume Breaks Through 30 Trillion Tokens! 🚀
ChatGPT Adds Automatic Memory Management, Priority Access for Paid Users
Google Upgrades Video Model Veo 3.1, Adds Audio Generation and Refined Editing
Anthropic Unveils Haiku 4.5 Model, Emphasizing Low Cost and High Performance
单词学习：![[cc14c38f-f822-463f-80ea-c5e93f830974.mp3]]
新闻播报：![[97bec2f7-df84-4cc1-918a-84337420df4d.mp3]]
通义千问上线“长记忆”功能，实现个性化连续对话
简报：
- 阿里云于10月16日宣布，旗下通义千问正式上线Qwen Chat Memory新功能。
- 该功能使AI助手具备“长记忆”能力，能存储、保留并主动回忆与用户的过往对话信息。
- 此功能旨在根据以往的互动经验，提供更具连贯性和个性化的回复。
- 其技术基础为通义团队此前发布的超万亿参数语言模型Qwen3-Max。
- ChatGPT、Gemini等其他AI产品此前也已推出类似记忆功能。
相关链接：
Tongyi Qianwen Launches "Long-Term Memory" Feature for Personalized Continuous Conversations 🧠
Brief:
- Alibaba Cloud announced on October 16 that its Tongyi Qianwen has officially launched the new Qwen Chat Memory feature.
- This feature grants the AI assistant "long-term memory" capability, enabling it to store, retain, and actively recall past conversation information with users. 💡
- The function aims to provide more coherent and personalized responses based on previous interaction experiences.
- Its technical foundation is the Qwen3-Max large language model with over a trillion parameters, previously released by the Tongyi team.
- Other AI products like ChatGPT and Gemini have also previously launched similar memory features. ✨
Related Links:
Modalities
/moʊˈdæl.ɪ.ti/
n. 方式，形态
▶ "The therapy involves multiple modalities to address the patient's needs."
[例句] 这种疗法涉及多种方式来满足患者的需求。
文心助手升级支持8种AIGC模态，可一句话生成3分钟视频
简报：
- 10月15日，百度宣布对文心助手进行重大升级，增强了其AIGC（人工智能生成内容）多模态创作与智能任务解决能力。
- 升级后，文心助手可支持AI图片、视频、音乐、播客等8种不同类型内容的创作。
- 用户现在只需输入一句话，即可生成长达3分钟的剧情短片，AI将自动完成情节设计、角色设定、配乐等全过程。
- 其自研“蒸汽机”模型突破了传统AI视频10秒时长的限制，实现了“无限时长”的实时交互生成。
- 此次升级还推出了行业首个开放式实时互动数字人智能体，具备高真实感形象和低延迟情感交互能力。
相关链接：
Wenxin Assistant Upgraded to Support 8 AIGC Modalities, Can Generate 3-Minute Videos from a Single Sentence 🚀
Brief:
- On October 15th, Baidu announced a major upgrade to Wenxin Assistant, significantly enhancing its AIGC (Artificial Intelligence Generated Content) multimodal creation and intelligent task-solving capabilities.
- Following the upgrade, Wenxin Assistant now supports the creation of 8 different types of content, including AI images, videos, music🎶, and podcasts.
- Users can now generate up to a 3-minute narrative short film by simply entering a single sentence. The AI will automatically handle the entire process, including plot design, character setup, and soundtrack. 🎬
- Its self-developed "Steam Engine" model breaks through the traditional 10-second limit for AI videos, enabling "unlimited duration" real-time interactive generation.
- This upgrade also introduces the industry's first open real-time interactive digital human intelligent agent, featuring highly realistic visuals and low-latency emotional interaction capabilities.
Related Link:
Explode
/ɪkˈsploʊd/
v. 爆炸
▶ "The bomb exploded in the crowded market, causing chaos."
[例句] 炸弹在拥挤的市场中爆炸，引发了混乱。
◼
衍生词
豆包大模型日均调用量突破30万亿Tokens
简报：
- 字节跳动旗下火山引擎于10月16日披露，豆包大模型的日均tokens使用量已超过30万亿，相较于2024年5月的1200亿tokens增长了253倍。
相关链接：
Doubao Large Model Daily Invocation Volume Breaks Through 30 Trillion Tokens! 🚀
Brief:
- ByteDance's Volcengine disclosed on October 16 that the Doubao large model's daily token usage has surpassed 30 trillion, marking a staggering 253-fold increase compared to 120 billion tokens in May 2024. 🤯 That's massive growth! ✨
Related Links:
Massive
/ˈmæs·ɪv/
adj. 巨大的
▶ "The project required a massive amount of resources to complete."
[例句] 这个项目需要大量资源才能完成。
ChatGPT新增自动记忆管理，付费用户优先体验
简报：
- OpenAI 为 ChatGPT 推出了全新的自动记忆管理功能，旨在解决长期使用中记忆空间配额已满的问题。
- 启用后，系统会根据时间顺序和用户设定的优先级，自动删除旧的或不重要的记忆，避免存储溢出。
- 用户可以手动调整记忆的优先级，以长期保留重要信息。
- 该功能目前仅向 ChatGPT Plus 和 Pro 的付费订阅用户开放，并已于2025年10月16日起陆续推送。
相关链接：
ChatGPT Adds Automatic Memory Management, Priority Access for Paid Users
Brief:
- OpenAI has rolled out a new automatic memory management feature for ChatGPT, designed to address the issue of memory quota being full during long-term use. 🤖
- Once enabled, the system will automatically delete old or less important memories based on chronological order and user-defined priorities, preventing storage overflow. 🧹
- Users can manually adjust the priority of memories to retain important information long-term.
- This feature is currently available exclusively to paid subscribers of ChatGPT Plus and Pro, with a gradual rollout commencing on October 16, 2025. ✨
Related links:
Chronological
/ˌkrɒn.əˈlɒdʒ.ɪ.kəl/
adj. 按时间顺序的
▶ "The events were listed in chronological order in the report."
[例句] 报告中按时间顺序列出了事件。
谷歌升级视频模型Veo 3.1，新增音频生成与精细化编辑
简报：
- 谷歌发布其视频生成模型Veo的升级版Veo 3.1，新版本在音频输出、编辑控制粒度和图像转视频质量方面实现了改进。
- Veo 3.1的核心升级在于为所有编辑功能增加了音频生成能力，并允许用户向视频中添加新对象，系统会自动将其融入原有画面。
- 该模型将被集成到视频编辑器Flow、Gemini应用程序以及面向开发者的Vertex AI和Gemini API接口中。
相关链接：
Google Upgrades Video Model Veo 3.1, Adds Audio Generation and Refined Editing
Brief:
- Google has released an upgraded version of its video generation model, Veo 3.1. The new version features improvements in audio output, granular editing control, and image-to-video quality. 🚀
- A core upgrade in Veo 3.1 is the addition of audio generation capabilities for all editing functions, and it allows users to add new objects to videos, which the system automatically integrates into the existing footage. 🔊
- The model will be integrated into the video editor Flow, the Gemini application, and for developers via the Vertex AI and Gemini API interfaces. 🔗
Related Links:
Granular
/ˈɡræn.jə.lɚ/
adj. 颗粒状的；细致的
▶ "The report provides a granular analysis of the market trends."
[例句] 该报告提供了市场趋势的细致分析。
◼
衍生词
granularity (n.)
颗粒度；细致程度
Anthropic发布Haiku 4.5模型，主打低成本与高性能
简报：
- 10月16日，人工智能公司Anthropic发布了新款小型AI模型Claude Haiku 4.5，专为聊天助手、客户服务等需要即时反馈的实时任务设计。
- 该模型以三分之一的价格和超过两倍的速度，提供了与旧款大型模型Sonnet 4相当的编程能力。
- 在权威基准测试SWE-bench中，Haiku 4.5得分73.3%，略高于Sonnet 4的72.7%，部分基准测试表现接近OpenAI的GPT-5模型。
- API调用价格设定为每百万输入token 1美元，输出token 5美元，为开发者提供了高性价比的选择。
- Haiku 4.5已向所有Claude网页、App及API用户开放，并成为免费用户的默认模型。
相关链接：
Brief:
- On October 16, AI company Anthropic released its new compact AI model, Claude Haiku 4.5, specifically designed for real-time tasks requiring instant feedback, such as chatbots and customer service. 🚀
- The model offers programming capabilities comparable to the older, larger Sonnet 4 model, but at one-third the price and over twice the speed.
- In the authoritative SWE-bench benchmark test, Haiku 4.5 scored 73.3%, slightly higher than Sonnet 4's 72.7%, with some benchmarks showing performance close to OpenAI's GPT-5 model.
- API call pricing is set at dJyH6kDkfC5RlM16.answerText per million input tokens and $5 per million output tokens, providing developers with a highly cost-effective option. 💸
- Haiku 4.5 is now available to all Claude web, app, and API users, and has become the default model for free users. ✨
Related links:
Skyrocket
/ˈskaɪˌrɑː.kɪt/
v. 急升，猛涨
▶ "The company's profits skyrocketed after the new product launch."
[例句] 新产品发布后，公司利润猛涨。
本文作者:topwind
本文链接:
版权声明:本博客所有文章除特别声明外，均采用 BY-NC-SA
许可协议。转载请注明出处！
