AI Industry Daily: Insights into the Frontier, Grasping the Future

July 7, 2025

Today, the field of artificial intelligence continues to see new breakthroughs and applications, from the iterative upgrades of large model technologies to the implementation of specific industry solutions, all demonstrating the vigorous development of AI technology. This report aims to summarize recent hot events in the AI industry and provide readers with a comprehensive and in-depth overview of industry dynamics.

Hot Topics Overview

Recently, the AI industry has shown strong innovation vitality in multiple dimensions. At the technical level, large language models and multimodal AI continue to evolve, with embodied intelligence and AI Agents becoming new focal points. At the application level, AI is deeply integrated into social, design, video generation, and other fields, enhancing user experience and industry efficiency. Meanwhile, the capital market's attention to AI remains undiminished, with frequent financing activities for related companies, indicating market confidence in the future development of AI. It is worth noting that while AI improves efficiency, it also raises deeper considerations regarding data privacy, ethics, and human-computer collaboration models, which require joint attention and exploration of solutions.

Specific Hot Topics

1. Technological Innovation and Model Breakthroughs

ByteDance Open-Sources AI IDE Core Component Trae-Agent: ByteDance has launched Trae-Agent, an intelligent assistant based on large language models, specifically designed for software engineering tasks. It can independently perform code understanding, bug reproduction, solution formulation, and high-quality code writing. Trae-Agent supports various language models, including OpenAI, and integrates file editing and script execution functions, while also having the ability to automatically save operation logs, improving the transparency and debugging convenience of the development process. This marks a further penetration of AI in the field of software development, which is expected to significantly improve development efficiency.
Zhipu AI Releases and Open-Sources GLM-4.1V-Thinking Series Visual Models: Zhipu AI has made significant progress in the AI field by open-sourcing the new generation of general visual language model GLM-4.1V-Thinking. This model possesses multimodal input capabilities for images, videos, and documents, and has demonstrated excellent performance in multiple authoritative evaluations, especially in complex reasoning tasks. In addition, Zhipu AI has launched the MaaS "Agent Application Space" platform, aiming to reduce the threshold for enterprises to access Agent technology through special support programs, and promote the development of an AI-native entrepreneurial ecosystem. This indicates that multimodal AI and Agent technology are becoming new trends in AI development.
Baidu Launches Self-Developed Multimodal Large Model MuseSteamer and AI Video Creation Platform: Baidu has released its self-developed video generation model MuseSteamer and its accompanying AI video creation platform. MuseSteamer is the world's first model to achieve integrated Chinese audio and video generation, breaking the traditional AIGC video production process of "first picture, then sound." It can achieve collaborative creation of visuals, sound effects, and human voice dialogue. Users only need to upload one image to generate professional-grade video content. This innovation will greatly simplify the video production process, lower the creation threshold, and bring convenience to content creators.
Google Veo 3 AI Text-to-Video Model Officially Opens to Pro/Ultra Members: Google's latest generation AI text-to-video model, Veo 3, has been officially opened to Google AI Pro and Ultra members. This model supports generating 1080p high-definition videos, with internal tests reaching 4K resolution, offering rich and realistic visual details. Veo 3 is the first model to support synchronized video and audio generation, automatically generating environmental sound effects, character dialogues, and background music. It also supports text or image input for video generation, suitable for complex prompt instructions and multi-shot narratives, improving creation efficiency. In the future, Veo 3 will add a "photo-to-video" function, further expanding its application scenarios.
Kunlun Tech Open-Sources Reward Model Skywork-Reward-V2 Again: Kunlun Tech has open-sourced the second generation of its reward model, Skywork-Reward-V2 series, which includes 8 models with different parameter scales (from 600 million to 8 billion) and has achieved top performance in multiple mainstream evaluation benchmarks. This series is built on high-quality mixed datasets, demonstrating strong generalization and practical capabilities. This move will further promote the development of AI model training and optimization, providing stronger basic support for AI applications.
OmniGen2 Undergoes Major Upgrade, Unifying Image Generation for Further Evolution: Zhipu AI announced a significant upgrade to its OmniGen2 image generation model. OmniGen2 adopts a decoupled architecture and dual-encoder strategy, enhancing contextual understanding and instruction following capabilities, and greatly improving image generation quality. By restructuring the data generation process, it addresses issues with open-source dataset defects and introduces an image generation feedback mechanism to enhance the model's self-optimization capabilities. This indicates that image generation technology is moving towards higher quality and more intelligent directions.
Open-Source Revolution! Kyutai TTS Released: Ultra-Low Latency Speech Synthesis, Ushering in a New Era of AI Voice!: The release of Kyutai TTS marks a new stage in open-source AI voice technology. This model supports streaming text transmission with a latency as low as 350 milliseconds, significantly improving the real-time voice interaction experience. Its speech generation accuracy is high, with word error rates for English and French as low as 2.82% and 3.29% respectively, and it also supports word timestamp output. The open-source nature of Kyutai TTS allows for free use, modification, and distribution, which will promote innovation and technological progress in voice interaction within the global AI community.

2. Industry Applications and Business Model Innovation

JD.com Internally Tests "Pet TA" and "Healing Universe" AI Design Products: JD.com's app has quietly launched two AI social products: "Pet TA" and "Healing Universe." "Pet TA" provides companionship, dress-up, consultation, and one-click food purchase services centered around pet digital humans; "Healing Universe" combines emotional recognition, memory calendars, and community interaction with professional psychological counseling services. This indicates that AI is increasingly integrated into social and emotional companionship fields, meeting diverse user needs.
Tencent Yuanbao Supports One-Sentence Search for Images and Video Content: Tencent Yuanbao has launched a new feature that allows users to match images and video account content with a "one-sentence search." After enabling "network search," Yuanbao can automatically match images and video accounts based on queries, supporting any model and not limited by whether "deep thinking" is enabled. This feature greatly improves the efficiency and convenience of information retrieval, providing users with a more intuitive and efficient way to obtain information.
WeChat Pay MCP Launched: Perfect Integration of AI and Payment, Ushering in a New Era of Business: The launch of WeChat Pay MCP brings new possibilities for AI commercialization. This feature provides new revenue channels for AI applications, allowing users to directly obtain services through payment. MCP builds a data closed-loop, enabling merchants to adjust service content and pricing in real-time to optimize ROI. Transaction data becomes a source for AI service optimization, enhancing user lifetime value and creating more profit opportunities. This heralds the deep integration of AI in the financial payment sector and the innovation of business models.
Meitu WHEE Launches "One-Sentence Image Editing" Function: WHEE's "one-sentence image editing" function allows users to perform complex image editing operations with simple voice commands, greatly enhancing the user experience. This function supports various style switches, such as futuristic and nostalgic artistic styles, and can add or remove text, accurately processing text content in photos. This makes image editing more convenient and intelligent, lowering the barrier to professional image editing.
Xingliu Agent Launched! A One-Stop Creative Design Agent More Suitable for Chinese Designers: Xingliu Agent has been officially launched as a one-stop creative design agent specifically designed for Chinese designers. It inherits Lovart's full-stack intelligent design capabilities, fully adapting to Chinese semantics, oriental aesthetics, and local scenarios. Users only need to input one sentence to automatically decompose tasks, match styles, and generate complete design materials, supporting multimodal content creation, including images, videos, and 3D formats. This provides designers with powerful AI-assisted tools, improving design efficiency and creative realization capabilities.

3. Market Dynamics and Capital Attention

Zhipu AI Receives 1 Billion Yuan Strategic Investment from Shanghai State-Owned Assets: Zhipu AI, a domestic AI large model enterprise, announced at the Open Platform Industry Ecosystem Conference that it has received a 1 billion yuan strategic investment from Shanghai state-owned assets, with the first batch of transactions completed by Pudong Venture Capital Group and Zhangjiang Group. At the same time, the three parties will cooperate with Shanghai Electric and Pudong Development Group to jointly build new AI infrastructure. This investment not only provides Zhipu AI with sufficient financial support but also reflects the continued optimism and strategic layout of state-owned capital in the AI large model field.
Figma Plans to Go Public on NYSE with a Valuation of Approximately $20 Billion, AI Design Has a Promising Future: Figma plans to go public on the NYSE with a valuation of approximately $20 billion, making it one of the most anticipated tech IPOs in 2025. Its strong financial performance (revenue of $749 million in 2024 and $1.54 billion in cash reserves) and proactive strategy in AI technology (launching tools like Figma Make, and integrating generative AI to optimize design workflows in the future) all indicate its huge potential in the AI design field. This shows that the capital market highly recognizes the value of AI-powered design tools.
Ambiq Micro, a Chip Design Company, Applies for US IPO, Benefiting from Generative AI-Driven Market Demand: Ambiq Micro achieved a 16.1% net sales growth in 2024. Although still in a loss-making state, its technological advantages in ultra-low-power semiconductors have given it a favorable position in the edge AI market. The company plans to raise funds through the IPO for product development and market expansion. This reflects the strong driving effect of generative AI on the chip industry and the urgent market demand for high-efficiency AI chips.
Perplexity Max Subscription Launched, Monthly Fee $200: Perplexity has launched its premium subscription service, Perplexity Max, for $200 per month (approximately 1433 RMB). Subscribers can access Labs, a spreadsheet and report generation tool, without restrictions, and experience new features like the Comet browser in advance, while also being able to call advanced AI models such as OpenAI's gpt-3 pro and Claude Opus 4. This indicates that AI products are exploring high-end paid models to provide more professional and powerful services.

Conclusion

In summary, the current AI industry is in a phase of rapid development and deep integration. Technological innovations continue to break boundaries, especially in the fields of large models, multimodal AI, and AI Agents, showing huge potential and application prospects. AI technology is accelerating its penetration into various industries, giving rise to new application scenarios and business models, greatly improving production efficiency and user experience. At the same time, the continuous investment of capital in the AI field also provides a solid foundation for the healthy development of the industry. However, with the widespread application of AI, issues such as data security, ethical norms, and human-computer collaboration are becoming increasingly prominent, requiring joint attention and exploration of solutions from within and outside the industry. In the future, AI will continue to develop towards a more intelligent, more inclusive, and more responsible direction, profoundly changing our work and life.

Hot Topics Overview

Specific Hot Topics

1. Technological Innovation and Model Breakthroughs

ByteDance Open-Sources AI IDE Core Component Trae-Agent: ByteDance has launched Trae-Agent, an intelligent assistant based on large language models, specifically designed for software engineering tasks. It can independently perform code understanding, bug reproduction, solution formulation, and high-quality code writing. Trae-Agent supports various language models, including OpenAI, and integrates file editing and script execution functions, while also having the ability to automatically save operation logs, improving the transparency and debugging convenience of the development process. This marks a further penetration of AI in the field of software development, which is expected to significantly improve development efficiency.

Zhipu AI Releases and Open-Sources GLM-4.1V-Thinking Series Visual Models: Zhipu AI has made significant progress in the AI field by open-sourcing the new generation of general visual language model GLM-4.1V-Thinking. This model possesses multimodal input capabilities for images, videos, and documents, and has demonstrated excellent performance in multiple authoritative evaluations, especially in complex reasoning tasks. In addition, Zhipu AI has launched the MaaS "Agent Application Space" platform, aiming to reduce the threshold for enterprises to access Agent technology through special support programs, and promote the development of an AI-native entrepreneurial ecosystem. This indicates that multimodal AI and Agent technology are becoming new trends in AI development.

Baidu Launches Self-Developed Multimodal Large Model MuseSteamer and AI Video Creation Platform: Baidu has released its self-developed video generation model MuseSteamer and its accompanying AI video creation platform. MuseSteamer is the world's first model to achieve integrated Chinese audio and video generation, breaking the traditional AIGC video production process of "first picture, then sound." It can achieve collaborative creation of visuals, sound effects, and human voice dialogue. Users only need to upload one image to generate professional-grade video content. This innovation will greatly simplify the video production process, lower the creation threshold, and bring convenience to content creators.

Google Veo 3 AI Text-to-Video Model Officially Opens to Pro/Ultra Members: Google's latest generation AI text-to-video model, Veo 3, has been officially opened to Google AI Pro and Ultra members. This model supports generating 1080p high-definition videos, with internal tests reaching 4K resolution, offering rich and realistic visual details. Veo 3 is the first model to support synchronized video and audio generation, automatically generating environmental sound effects, character dialogues, and background music. It also supports text or image input for video generation, suitable for complex prompt instructions and multi-shot narratives, improving creation efficiency. In the future, Veo 3 will add a "photo-to-video" function, further expanding its application scenarios.

Kunlun Tech Open-Sources Reward Model Skywork-Reward-V2 Again: Kunlun Tech has open-sourced the second generation of its reward model, Skywork-Reward-V2 series, which includes 8 models with different parameter scales (from 600 million to 8 billion) and has achieved top performance in multiple mainstream evaluation benchmarks. This series is built on high-quality mixed datasets, demonstrating strong generalization and practical capabilities. This move will further promote the development of AI model training and optimization, providing stronger basic support for AI applications.

OmniGen2 Undergoes Major Upgrade, Unifying Image Generation for Further Evolution: Zhipu AI announced a significant upgrade to its OmniGen2 image generation model. OmniGen2 adopts a decoupled architecture and dual-encoder strategy, enhancing contextual understanding and instruction following capabilities, and greatly improving image generation quality. By restructuring the data generation process, it addresses issues with open-source dataset defects and introduces an image generation feedback mechanism to enhance the model's self-optimization capabilities. This indicates that image generation technology is moving towards higher quality and more intelligent directions.

Open-Source Revolution! Kyutai TTS Released: Ultra-Low Latency Speech Synthesis, Ushering in a New Era of AI Voice!: The release of Kyutai TTS marks a new stage in open-source AI voice technology. This model supports streaming text transmission with a latency as low as 350 milliseconds, significantly improving the real-time voice interaction experience. Its speech generation accuracy is high, with word error rates for English and French as low as 2.82% and 3.29% respectively, and it also supports word timestamp output. The open-source nature of Kyutai TTS allows for free use, modification, and distribution, which will promote innovation and technological progress in voice interaction within the global AI community.

2. Industry Applications and Business Model Innovation

JD.com Internally Tests "Pet TA" and "Healing Universe" AI Design Products: JD.com's app has quietly launched two AI social products: "Pet TA" and "Healing Universe." "Pet TA" provides companionship, dress-up, consultation, and one-click food purchase services centered around pet digital humans; "Healing Universe" combines emotional recognition, memory calendars, and community interaction with professional psychological counseling services. This indicates that AI is increasingly integrated into social and emotional companionship fields, meeting diverse user needs.

Tencent Yuanbao Supports One-Sentence Search for Images and Video Content: Tencent Yuanbao has launched a new feature that allows users to match images and video account content with a "one-sentence search." After enabling "network search," Yuanbao can automatically match images and video accounts based on queries, supporting any model and not limited by whether "deep thinking" is enabled. This feature greatly improves the efficiency and convenience of information retrieval, providing users with a more intuitive and efficient way to obtain information.

WeChat Pay MCP Launched: Perfect Integration of AI and Payment, Ushering in a New Era of Business: The launch of WeChat Pay MCP brings new possibilities for AI commercialization. This feature provides new revenue channels for AI applications, allowing users to directly obtain services through payment. MCP builds a data closed-loop, enabling merchants to adjust service content and pricing in real-time to optimize ROI. Transaction data becomes a source for AI service optimization, enhancing user lifetime value and creating more profit opportunities. This heralds the deep integration of AI in the financial payment sector and the innovation of business models.

Meitu WHEE Launches "One-Sentence Image Editing" Function: WHEE's "one-sentence image editing" function allows users to perform complex image editing operations with simple voice commands, greatly enhancing the user experience. This function supports various style switches, such as futuristic and nostalgic artistic styles, and can add or remove text, accurately processing text content in photos. This makes image editing more convenient and intelligent, lowering the barrier to professional image editing.

Xingliu Agent Launched! A One-Stop Creative Design Agent More Suitable for Chinese Designers: Xingliu Agent has been officially launched as a one-stop creative design agent specifically designed for Chinese designers. It inherits Lovart's full-stack intelligent design capabilities, fully adapting to Chinese semantics, oriental aesthetics, and local scenarios. Users only need to input one sentence to automatically decompose tasks, match styles, and generate complete design materials, supporting multimodal content creation, including images, videos, and 3D formats. This provides designers with powerful AI-assisted tools, improving design efficiency and creative realization capabilities.

3. Market Dynamics and Capital Attention

Zhipu AI Receives 1 Billion Yuan Strategic Investment from Shanghai State-Owned Assets: Zhipu AI, a domestic AI large model enterprise, announced at the Open Platform Industry Ecosystem Conference that it has received a 1 billion yuan strategic investment from Shanghai state-owned assets, with the first batch of transactions completed by Pudong Venture Capital Group and Zhangjiang Group. At the same time, the three parties will cooperate with Shanghai Electric and Pudong Development Group to jointly build new AI infrastructure. This investment not only provides Zhipu AI with sufficient financial support but also reflects the continued optimism and strategic layout of state-owned capital in the AI large model field.

Figma Plans to Go Public on NYSE with a Valuation of Approximately $20 Billion, AI Design Has a Promising Future: Figma plans to go public on the NYSE with a valuation of approximately $20 billion, making it one of the most anticipated tech IPOs in 2025. Its strong financial performance (revenue of $749 million in 2024 and $1.54 billion in cash reserves) and proactive strategy in AI technology (launching tools like Figma Make, and integrating generative AI to optimize design workflows in the future) all indicate its huge potential in the AI design field. This shows that the capital market highly recognizes the value of AI-powered design tools.

Ambiq Micro, a Chip Design Company, Applies for US IPO, Benefiting from Generative AI-Driven Market Demand: Ambiq Micro achieved a 16.1% net sales growth in 2024. Although still in a loss-making state, its technological advantages in ultra-low-power semiconductors have given it a favorable position in the edge AI market. The company plans to raise funds through the IPO for product development and market expansion. This reflects the strong driving effect of generative AI on the chip industry and the urgent market demand for high-efficiency AI chips.

Perplexity Max Subscription Launched, Monthly Fee $200: Perplexity has launched its premium subscription service, Perplexity Max, for $200 per month (approximately 1433 RMB). Subscribers can access Labs, a spreadsheet and report generation tool, without restrictions, and experience new features like the Comet browser in advance, while also being able to call advanced AI models such as OpenAI's gpt-3 pro and Claude Opus 4. This indicates that AI products are exploring high-end paid models to provide more professional and powerful services.

Conclusion