![[77261bce-2c7b-4ca3-a04b-0180455f0947.mp3]]

通义千问发布Qwen VLo多模态模型，提升图像理解与创作能力

简报：

• 通义千问近期正式发布Qwen VLo多模态大模型，在图像内容理解与高质量生成方面实现突破，支持渐进式图片生成与开放指令编辑，提升视觉创作体验。
• Qwen VLo增强了语义一致性和结构保留能力， 支持多语言指令输入及文本生成图片，可广泛应用于内容创作、图像修改和多模态理解等场景。
• 用户可在Qwen Chat平台（chat.qwen.ai）体验该模型，目前处于预览阶段，研发团队将持续优化性能。

相关链接：

• AIBase新闻报道1

Tongyi Qianwen Releases Qwen VLo Multimodal Model, Enhancing Image Understanding and Creation Capabilities

Briefing:

• Tongyi Qianwen recently officially launched the Qwen VLo multimodal large model, achieving breakthroughs in image content understanding and high-quality generation. It supports progressive image generation and open-ended instruction editing, enhancing the visual creation experience. ✨
• Qwen VLo enhances semantic consistency and structure preservation capabilities, supporting multi-language instruction input and text-to-image generation. It can be widely applied in scenarios such as content creation, image modification, and multimodal understanding. 🎨
• Users can experience the model on the Qwen Chat platform (chat.qwen.ai). It is currently in a preview phase, and the R&D team will continue to optimize its performance. 🚀

Related Links:

• AIBase News Report 1

OpenAI推出高阶自动化研究API，支持网页搜索和多任务分析

简报：

• OpenAI于2025年6月26日发布两款全新Deep Research API模型：o3-deep-research-2025-06-26和o4-mini-deep-research-2025-06-26，专为高阶分析、深度信息合成及自动网页搜索、数据分析和代码执行等场景设计。
• o3模型主打精密推理和高准确率，适用于复杂科研和金融分析，o4-mini则兼顾效率和成本，适合大规模快速查询，价格分别为每千次调用10至40美元和2至8美元。
• Deep Research API跳过ChatGPT的交互澄清步骤，直接生成结构化、带引用的报告，适合市场分析、学术研究等场景，同时支持异步和Webhook通知以提升开发效率。
• OpenAI强调API当前只限于ChatGPT生态内部测试，以评估“现实世界说服风险”，未来计划进一步优化模型功能并探索更广泛的数据源整合，面对谷歌、DeepSeek等竞争对手，开放API以巩固企业级市场地位。

相关链接：

• OpenAI Deep Research API发布首发解读
• OpenAI两大深度研究模型API上线：功能、定价与应用场景全览
• Deep Research API与开源竞品对比分析

OpenAI Launches Advanced Automated Research API, Supporting Web Search and Multi-task Analysis

Brief:

• On June 26, 2025, OpenAI released two new Deep Research API models: o3-deep-research-2025-06-26 and o4-mini-deep-research-2025-06-26. These models are designed for advanced analysis, deep information synthesis, and scenarios involving automated web search, data analysis, and code execution. 🚀
• The o3 model excels in precise reasoning and high accuracy, suitable for complex scientific research and financial analysis. The o4-mini balances efficiency and cost, ideal for large-scale rapid queries. Pricing is set at $10-40 and$2-8 per thousand calls, respectively. 💰
• The Deep Research API bypasses ChatGPT's interactive clarification steps, directly generating structured, cited reports suitable for market analysis, academic research, and similar applications. It also supports asynchronous operations and Webhook notifications to enhance development efficiency.
• OpenAI emphasizes that the API is currently limited to internal testing within the ChatGPT ecosystem to evaluate "real-world persuasion risks." Future plans include further optimizing model functionalities and exploring broader data source integration. By opening up the API, OpenAI aims to solidify its position in the enterprise market amidst competitors like Google and DeepSeek. 💡

Related Links:

• OpenAI Deep Research API发布首发解读
• OpenAI两大深度研究模型API上线：功能、定价与应用场景全览
• Deep Research API与开源竞品对比分析

可灵AI上线同步生成功能，视频音效高度还原画面内容

简报：

• 可灵AI宣布其全系列视频模型正式上线“视频音效”功能，用户在生成视频的同时可自动获得与画面内容同步的立体声音效，实现“所见即所听”的沉浸体验。
• 新升级的“音效生成”功能增添“视频生音效”模块，支持用户上传视频或调用历史作品，一键匹配合适音效，利用自研模型 Kling-Foley 实现音画帧级对齐。
• 该功能目前已面向所有用户限时免费开放，增强了AI驱动的视听融合体验。

相关链接：

• 可灵AI上线“所见即所听”视频音效功能，实现沉浸式体验
• 可灵AI Kling-Foley 多模态模型音效升级

Keling AI Launches Synchronous Generation, Video Sound Effects Faithfully Recreate On-Screen Content 🔊🎥✨

Briefing:

• Keling AI announced that its full suite of video models officially incorporates a "Video Sound Effects" feature. Users can now automatically generate synchronized stereo sound effects that perfectly match the visual content while creating videos, delivering an immersive "what you see is what you hear" experience.
• The newly upgraded "Sound Effect Generation" function introduces a "Video to Sound Effect" module. This allows users to upload videos or leverage their past creations, enabling one-click matching of appropriate sound effects by utilizing Keling AI's self-developed Kling-Foley model for frame-level audio-visual alignment.
• This feature is currently available to all users for a limited-time free trial, significantly enhancing the AI-driven audio-visual fusion experience.

Related Links:

• Keling AI Launches "What You See Is What You Hear" Video Sound Effect Feature, Achieving Immersive Experience
• Keling AI Kling-Foley Multimodal Model Sound Effect Upgrade

Hengbot推出多功能AI机器狗Sirius，集成语音互动与自定义能力

简报：

• Hengbot公司正式发布了Sirius机器狗，这款宠物机器人具备跳舞、踢球及语音交流等多项功能，并集成OpenAI大语言模型支持AI陪聊。
• Sirius拥有14个运动轴和专属Neurocore关节系统，使动作更加自然流畅，可根据用户设定实现多种自定义外观和声音，并支持Python、C、C++等编程扩展。
• 机器狗重量约1公斤，采用航空级铝合金材质，最适合于平坦室内活动，续航运动时为40~60分钟，静止待机可达2小时，充电时长约1小时。
• 该产品官方承诺不收集用户数据，目前已开启预售，预售价为1299美元，预计2025年秋季上市。

相关链接：

• AI Base资讯
• IT之家报道
• 三立新闻网

Hengbot Unveils Versatile AI Robot Dog Sirius, Integrating Voice Interaction and Customization Capabilities

Briefing:

• Hengbot has officially launched the Sirius robot dog 🐶. This pet robot offers multiple functions including dancing, kicking balls, and voice interaction, powered by an integrated OpenAI large language model for AI companionship.
• Sirius features 14 movement axes and an exclusive Neurocore joint system, enabling more natural and fluid movements ✨. It allows for various customizable appearances and sounds based on user settings, and supports programming extensions in Python, C, C++, and more.
• Weighing approximately 1 kg and crafted from aerospace-grade aluminum alloy, the robot dog is best suited for flat indoor activities. It offers a battery life of 40-60 minutes during active use, up to 2 hours in standby mode, and charges in about 1 hour.
• The product officially pledges not to collect user data 🛡️. It is currently available for pre-order at a pre-sale price of $1299, with an anticipated launch in Fall 2025.

Related Links:

• AI Base News
• IT Home Report
• Sanli News Network