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.
New AI products Click to learn more:https://app.aibase.com/zh
1. Qwen3.8-Omni-Flash launched: 1M context length, average performance improvement over 26% in 30 evaluations
Qwen launched a new multimodal model, Qwen3.8-Omni-Flash, which supports text, images, audio, and video input and has a 1M long context. Its performance has significantly improved in multiple evaluations. The model has enhanced audio and video processing capabilities and optimized API pricing. It also opens up related tools to support more complex workflows.
【AiBase Summary:】
🔥 Qwen3.8-Omni-Flash supports multimodal input and has a 1M long context capability
📈 Performance improved by over 26% on average in 30 evaluations, with audio and video processing capabilities close to Gemini3.8Flash
💰 API prices have dropped significantly, with audio and audio-video input costs reduced by 98% and 93%
2. Zhipu releases GLM-5.3-FlashX: speed reaches 200 tokens/s, domestic computing power speeds up again
Zhipu AI's BigModel platform released the GLM-5.3-FlashX large model, with a maximum output speed of 200 tokens/s, emphasizing the three-in-one competitiveness of "intelligence, price, and speed", providing high-throughput, low-latency inference services for enterprises and developers. The model further improves generation speed while maintaining its original level of intelligence and affordable pricing, reinforcing its commercial advantages in high-concurrency scenarios.
【AiBase Summary:】
🧠 GLM-5.3-FlashX large model has a maximum output speed of 200 tokens/s, improving inference efficiency.
💰 Maintains the original level of intelligence and affordable pricing, enhancing cost-effectiveness.
🚀 Supports commercial needs in high-concurrency scenarios, promoting the popularization of domestic large models.
3. Google's strongest Gemini 4 Pro tests SVG image generation impressively
Google is testing its strongest model, Gemini4Pro, under the name 'gemini-3.8-flash' on benchmark platforms such as Arena. Its internal code name is Argon. In the test version, Gemini4Pro performs well in generating SVG images, especially in understanding complex structures and geometric relationships. Netizens believe that after its official release, its capabilities may be further improved.
【AiBase Summary:】
✨ Google is testing its strongest model, Gemini4Pro, with the internal code name Argon.
🎨 Gemini4Pro performs well in SVG image generation, especially in handling complex structures and geometric relationships.
🚀 The test version may not be complete yet, but its capabilities may be further improved after the official release.
4. Apple AI Server Exposed: Code Name Baltra, Equipped with Self-developed M8 Ultra, May Use NVIDIA NVLink Fusion
Apple plans to return to the server market, developing enterprise-level AI servers, and collaborating with Broadcom to develop the AI server chip Baltra, while considering the use of NVIDIA NVLink Fusion technology to enhance performance.
【AiBase Summary:】
🍎 Apple is developing enterprise-level AI servers, possibly equipped with self-developed M8 Ultra chips.
🧩 Collaborating with Broadcom to develop the AI server chip Baltra, using TSMC's 3nm N3E process.
🔌 Apple considers using NVIDIA NVLink Fusion technology to connect M8 chips, enhancing AI computing capabilities.
5. Tesla Releases Update: Doubao Large Model Voice Assistant Expands to More Models, Dozens of Features Upgraded
Tesla released a new software update, introducing the Doubao large model voice assistant and expanding it to more models, while optimizing and upgrading multiple features to enhance user experience and security.
【AiBase Summary:】
🚗 Tesla released the 2026.26.200.11 software update, expanding the Doubao large model voice assistant to more models.
🔧 The update includes upgrades to multiple functions, including pet mode, blind spot warning, navigation, and dash cam.
📱 Added parent control function, optimized Apple Music operations, and fixed multiple security vulnerabilities.
6. Chen Hang, Founder of DingTalk, Steps Down as CEO and Joins: Salary of 100,000 RMB as Non-Executive Director at Baiwang Shares
Chen Hang left his position as CEO of DingTalk to join Baiwang Shares as a non-executive director, although his salary is relatively low, his experience in enterprise products and AI fields holds significant value for Baiwang Shares.
【AiBase Summary:】
✅ Chen Hang stepped down as CEO of DingTalk and joined Baiwang Shares' board.
💰 Chen Hang's salary is only 100,000 RMB, which is symbolic.
📈 Baiwang Shares focuses on the field of enterprise tax and AI, with a large number of major clients.
7. Tencent AI Voice Assistant Chatterfly Quietly Tests: Speak and Write Immediately, Data Stored Locally Only
Tencent's new AI voice assistant, Chatterfly, quietly started internal testing, supporting full-scenario AI typing, with strong intent recognition capabilities and supporting continuous learning of user input habits.
【AiBase Summary:】
📱 Supports multi-scenario AI typing, with strong intent recognition capabilities
💼 Built-in 6 scene Skills, including meeting minutes, work reports, etc.
🔒 Voice transcription records and audio files are stored locally on the user's computer
8. Humanoid Robot Booster K1 Exploration Edition Goes On Sale: JD Price 38,999 Yuan
Booster K1 Exploration Edition is a embodied intelligent robot aimed at developers, featuring a lightweight and compact hardware design and strong perception and interaction capabilities. It is equipped with the Doubao large model, supporting voice conversation and semantic understanding, and provides open APIs and OTA upgrade functions, offering developers a cost-effective experimental platform.
【AiBase Summary:】
🤖 Lightweight and compact design, height of about 95cm, weight of about 19.5kg, suitable for personal use.
📷 Equipped with dual-eye depth camera and ring microphone, achieving accurate perception and interaction.
🛠️ Provides open APIs and OTA upgrades, facilitating developers for customization and iteration.
