Welcome to the "AI Daily" section! This is your guide to exploring the world of artificial intelligence every day. Every day, we present you with the latest content in the AI field, focusing on developers to help you understand technology trends and learn about innovative AI product applications.
Fresh AI products Click to learn more:https://app.aibase.com/zh
1. Anthropic releases Claude Haiku 5.5: Cost reduced by 75%, inference intensity can be adjusted by yourself
Anthropic has launched Claude Haiku 5.5, the fastest and most capable small model from the company, designed for high-frequency and low-cost scenarios. The running cost of this model has been significantly reduced, and it introduces adjustable inference intensity, allowing users to flexibly adjust computing resources based on task requirements. In addition, Haiku 5.5 now supports multi-cloud platforms, making it easier for enterprises to integrate quickly.
【AiBase Summary:】
🧠 The inference intensity of Haiku 5.5 can be adjusted, allowing users to adjust the depth of thinking according to task requirements.
💰 The operating cost of Haiku 5.5 has been reduced by 75% compared to the previous generation, with input and output costs being only $0.1 and $0.5 per million tokens respectively.
☁️ Haiku 5.5 is now available on the Claude Platform, AWS, Google Cloud, and Microsoft Azure, making it easy for enterprises to connect to existing workflows.
2. WorkBuddy launches an independent file browser: Right-click on local files to directly call AI
Tencent WorkBuddy has launched an independent file browser, integrating AI capabilities directly into local file operations to enhance office efficiency and user experience.
【AiBase Summary:】
📎 WorkBuddy has added an independent file browser function that supports direct opening and editing of multiple file formats.
📝 Users can choose to set WorkBuddy as the default open method to achieve a more convenient AI-assisted editing experience.
🔄 After file modifications, it will automatically save or prompt users to refresh to the latest version, ensuring data security and consistency.
3. Google embeds AI search into mobile phones: EmbeddingGemma2 is open source, less than 600MB and can search images, videos, and audio offline
Google's EmbeddingGemma2 is a lightweight, multi-modal open-source embedding model that can run on local devices without relying on cloud services, significantly improving privacy protection and retrieval efficiency.
【AiBase Summary:】
📱 Google launched EmbeddingGemma2, the first native multi-modal open-source embedding model with 740M parameters, supporting semantic mapping of text, code, images, videos, and audio.
🔒 The model uses less than 600MB of memory and supports offline operation, protecting user privacy data from being uploaded to the cloud.
🔍 Supports cross-modal retrieval, covering over 100 languages, suitable for efficient search of local photo albums, video clips, files, and code repositories.
4. Meta's AI assistant Muse is now available on iPad platform, with over 6.6 million downloads in just one month
Meta's AI assistant Muse has rapidly expanded to the iPad platform, achieving cross-platform coverage within a month and enhancing its agent capabilities through new connectors and partners. At the same time, Meta is working to solve the issue of AI agents being misidentified, promoting the development of open technical standards.
【AiBase Summary:】
🤖 The launch of Muse AI assistant on the iPad platform marks rapid progress in its cross-platform expansion.
💼 Meta enhances Muse's agent capabilities through new connectors and partners.
🔒 Meta collaborates with industry partners to develop open technical standards to address issues of AI agents being misidentified.
5. Musk: Grok Bot no longer relies solely on in-house models, but selects the best models like Claude, Midjourney, and Suno based on tasks
Musk announced that Grok Bot will no longer rely solely on in-house models, but instead use multiple third-party APIs based on task requirements, achieving more efficient AI task processing.
【AiBase Summary:】
🤖 Grok Bot will automatically select the best backend model for tasks, including third-party APIs such as Claude Opus5.5, Midjourney, and Suno.
💻 Grok Bot is a persistent AI colleague that can run multi-step tasks on cloud computers without requiring continuous user interaction.
🔄 Musk's strategic shift moves Grok Bot from relying on in-house models to a task-oriented hybrid architecture of multiple models.
6. OpenAI fully pushes GPT-6: Chat window grows interactive interface, free users can also play with smart UI
OpenAI introduced GPT-6, bringing smart UI, optimized response speed, and improved information accuracy, significantly enhancing user experience.
【AiBase Summary:】
🧠 GPT-6 introduces a smart UI to improve the interactive experience.
⚡ Optimized response speed to enable real-time interaction.
🌐 Improved online search and information accuracy.
7. Manus' parent company completes another round of over $500 million financing, Tencent and Sequoia continue to invest
Manus' parent company recently completed another round of over $500 million financing, with a significant increase in valuation. This round was led by Boyu Capital and IDG Capital, with existing shareholders such as Tencent and Sequoia China continuing to invest, showing high recognition from the capital market for its technology route and commercialization prospects.
【AiBase Summary:】
💰 Manus' parent company has completed another round of over $500 million financing, with a significant increase in valuation.
🚀 The funding was led by Boyu Capital and IDG Capital, with Tencent and Sequoia China continuing to invest as existing shareholders.
🌐 The company plans to use the funds for technological research and development, team expansion, and global expansion.
8. Google adjusts Gemini subscription strategy: Free version retains Flash-Lite only from October 9, paid tiers further widen
Google recently updated the access rights and usage restrictions of the Gemini model. Starting from October 9, free users can only access the Gemini3.5-Flash-Lite version, while AI Plus subscribers cannot use the Gemini3.6Flash and higher versions. Google AI Pro and Ultra users can still use advanced models and deep thinking features. Additionally, Google introduced three workload modes—low, medium, and high—to deduct quotas in real-time based on task complexity. This marks Google's entry into a new stage of fine-grained operation of large model applications in terms of computing power economic efficiency.
【AiBase Summary:】
🚫 Google has adjusted the access rights of the Gemini model, allowing free users to only use the Flash-Lite version.
💰 Paid subscription users are restricted from using advanced models, while high-value users still have access to deep thinking options.
⚙️ Introduced low, medium, and high workload modes to deduct quotas based on task complexity.
