![[447021fa-b672-4091-a50f-42ddd256a851.mp3]]

阿里通义千问3向量模型Qwen3-Embedding成为HuggingFace开源榜单最受欢迎大模型

简报：

• 2024年6月11日，全球最大AI开源社区HuggingFace最新榜单显示，阿里巴巴最新开源的通义千问3向量模型Qwen3-Embedding登顶全球大模型榜首。
• 通义千问3自开源一个月以来下载量突破1250万，成为近期最受开发者欢迎的开源大模型。

相关链接：

• 36氪报道
• AIBASE新闻1

Alibaba's Tongyi Qianwen 3 Vector Model, Qwen3-Embedding, Becomes Most Popular Large Model on HuggingFace Open-Source Ranking 🚀

Brief:

• On June 11, 2024, the latest ranking from HuggingFace, the world's largest AI open-source community, showed that Alibaba's newly open-sourced Tongyi Qianwen 3 vector model, Qwen3-Embedding, topped the global large model list. 🏆
• Since its open-source release a month ago, Tongyi Qianwen 3 has surpassed 12.5 million downloads, becoming the most popular open-source large model among developers recently. 🎉

Related Links:

• 36Kr Report
• AIBASE News 1

OpenAI大幅调降o3模型价格并上线o3-pro高性能AI

简报：

• OpenAI宣布其o3模型输入费用由每百万令牌10美元降至2美元、输出费用由40美元降至8美元，整体降价80%，并对缓存输入进一步打折。
• 同时推出新一代高性能o3-pro模型，定位应对高可靠性和复杂任务，输入/输出价格分别为20美元和80美元。
• o3-pro现已对ChatGPT Pro和Team用户开放，并将逐步覆盖企业与教育客户，但运行速度较慢且成本更高。
• 此举被视为OpenAI对Google Gemini等竞品的回应，或引发行业AI模型价格战；o3的新低价有望推动更多创新应用落地。
• OpenAI还宣布开源模型发布时间延迟至夏末，新模型在推理能力上将对标业界领先者。

相关链接：

• OpenAI adds o3 Pro to ChatGPT and drops o3 price by 80 per cent, but open-source AI is delayed

OpenAI Drastically Cuts o3 Model Prices and Launches High-Performance o3-pro AI

Briefing:

• OpenAI announced a significant price reduction for its o3 model: input fees dropped from $10 to$2 per million tokens, and output fees from $40 to$8, an overall 80% decrease. Further discounts apply to cached inputs. 📉
• Simultaneously, a new generation high-performance model, o3-pro, has been launched, designed for high reliability and complex tasks, with input/output prices of $20 and$80 respectively. 🚀
• o3-pro is currently available to ChatGPT Pro and Team users and will gradually roll out to enterprise and education clients, though it runs slower and is more costly.
• This move is seen as OpenAI's response to competitors like Google Gemini and could trigger an industry AI model price war; the new low price for o3 is expected to drive more innovative applications. 💡
• OpenAI also announced a delay in its open-source model release until late summer, with the new model aiming to match industry leaders in inference capabilities.

Related Links:

• OpenAI adds o3 Pro to ChatGPT and drops o3 price by 80 per cent, but open-source AI is delayed

苹果设备端AI模型追平谷歌阿里，但服务器端仍落后OpenAI

简报：

• 苹果最新的AI模型已集成于Apple Intelligence，支持iOS和macOS等系统，其设备端模型无须联网也能运行，文本生成质量与谷歌、阿里巴巴的同类产品相当；
• 然而，苹果服务器端AI模型在多项评测中不敌OpenAI一年前的GPT-4o，在图像分析领域也逊色于Meta的Llama4Scout；
• 新的Apple On-Device模型拥有约30亿参数，能支持15种语言，适用于文本生成与分析等功能，但市场对个性化Siri升级的进展表现出不满，部分用户因此提起诉讼。

相关链接：

• aibase新闻1
• aibase新闻2
• SETN iOS26资讯

Apple's On-Device AI Models Catch Up with Google and Alibaba, But Server-Side Still Lags OpenAI 🍎

Brief:

• Apple's latest AI models, integrated into Apple Intelligence, support systems like iOS and macOS. Their on-device models can run without an internet connection, and their text generation quality is comparable to similar products from Google and Alibaba.
• However, Apple's server-side AI models fall short of OpenAI's GPT-4o from a year ago in several evaluations, and are also inferior to Meta's Llama4Scout in the field of image analysis. 😔
• The new Apple On-Device model boasts approximately 3 billion parameters and supports 15 languages, suitable for functions like text generation and analysis. Yet, the market has expressed dissatisfaction with the progress of personalized Siri upgrades, leading some users to file lawsuits. ⚖️

Related Links:

• aibase news 1
• aibase news 2
• SETN iOS26 News

高通推新款骁龙AR1+ Gen 1处理器，智能眼镜支持本地AI

简报：

• 高通发布全新骁龙 AR1+ Gen 1 处理器，专为智能眼镜设计，可实现设备端AI运算，无需依赖手机或云端。
• 新处理器面积缩小28%，支架高度降低20%，功耗实现7%下降，提升续航能力和佩戴体验。
• 支持设备端个性化AI功能，加强数据隐私保护，并与Meta合作推动轻薄AI+AR眼镜的发展。
• 高通预计该芯片将推动更多集成人工智能的智能眼镜产品上市，为用户带来更优质的视觉AI体验。

相关链接：

• IT之家报道
• AIBase新闻

Qualcomm Launches New Snapdragon AR1+ Gen 1 Processor, Bringing On-Device AI to Smart Glasses

Briefing:

• Qualcomm has unveiled the new Snapdragon AR1+ Gen 1 processor, specifically designed for smart glasses to enable on-device AI computing without relying on a smartphone or cloud. 👓
• The new processor boasts a 28% smaller area, 20% lower bracket height, and a 7% reduction in power consumption, enhancing battery life and wearability.
• It supports personalized on-device AI features, strengthens data privacy protection, and partners with Meta to advance the development of thin and light AI+AR glasses.
• Qualcomm anticipates this chip will drive the launch of more AI-integrated smart glasses, offering users a superior visual AI experience. ✨🚀

Related Links:

• IT之家报道
• AIBase新闻

Figma推出官方MCP服务实现AI驱动设计到代码无缝转换

简报：

• 2025年6月，Figma正式上线Model Context Protocol（MCP）服务，通过简化、高效的数据传输方式，实现从设计到代码的自动生成，提升开发效率和设计一致性。
• 官方MCP支持元数据提取、高层次设计截图、交互行为伪代码和实时设计数据传输，可与主流IDE和AI编码工具直接集成，减少设计与开发沟通成本。
• MCP服务发布后，业界反响积极，设计师和开发者认为其有效解决了非官方工具数据不完整的问题，并期待未来支持远程服务器和更复杂设计系统。
• 该服务有望推动AI与创意产业深度融合，为跨职能团队提供更智能、更高效的协同工具，促进行业标准化发展。

相关链接：

• AIbase新闻报道
• 36氪简讯
• IT之家新闻

Figma Launches Official MCP Service for Seamless AI-Driven Design-to-Code Conversion

Briefing:

• In June 2025, Figma officially launched its Model Context Protocol (MCP) service, enabling automatic generation from design to code through a simplified and efficient data transfer method, thereby boosting development efficiency and design consistency. 🚀
• The official MCP supports metadata extraction, high-level design screenshots, interactive behavior pseudocode, and real-time design data transmission. It can be directly integrated with mainstream IDEs and AI coding tools, reducing communication costs between design and development.
• Following the MCP service launch, industry feedback has been positive. Designers and developers believe it effectively resolves the issue of incomplete data from unofficial tools and look forward to future support for remote servers and more complex design systems. ✨
• This service is expected to foster deeper integration between AI and the creative industry, providing smarter and more efficient collaboration tools for cross-functional teams, and promoting industry standardization. 💡

Related Links:

• AIbase News Report
• 36Kr Newsflash
• IT Home News