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 technical trends and learn about innovative AI product applications.

Fresh AI products Click to learn more:https://app.aibase.com/zh

1. Baidu's ERNIE-5.0-0110 officially released, ranking second in mathematical capabilities globally

Baidu's latest ERNIE large model, ERNIE-5.0-0110, has shown excellent performance in multiple authoritative evaluations, especially in mathematical processing capabilities, which rank second globally. The model has also achieved significant enhancements in programming development, expert knowledge reserves, and creative writing, enabling it to execute complex instructions more accurately.

image.png

AiBase Highlights:

🧠 ERNIE-5.0-0110 ranks second in mathematical processing capabilities globally, second only to GPT-5.2-High.

💻 The model has significant improvements in programming development, expert knowledge reserves, and creative writing.

🏥 It demonstrates high professional standards in multiple specialized fields such as scientific research, business finance, and healthcare.

Details: https://ernie.baidu.com/

2. Step-Audio-R1.1 from StepXingchen ranks first globally

The open-source speech inference model Step-Audio-R1.1 launched by StepXingchen Company has achieved first place in global artificial intelligence model evaluations, demonstrating its outstanding capabilities in audio processing and logical reasoning. The model has deep speech reasoning capabilities and real-time response features, allowing end-to-end understanding of speech content and supporting streaming inference. Users can download the model for free through HuggingFace and try it on the open platform.

image.png

AiBase Highlights:

🌟 Step-Audio-R1.1 ranked first globally with an accuracy rate of 96.4% in international evaluations!

📈 The model has deep speech reasoning and real-time response capabilities, supporting streaming inference functions.

💻 Users can download the model for free through HuggingFace and experience it on the open platform.

Details: https://huggingface.co/stepfun-ai/Step-Audio-R1.1

3. Qwen fully integrated into Alibaba's ecosystem: order food, buy tickets with one sentence, starting the "AI Handling Affairs" era

The Qwen App has fully integrated into Alibaba's core ecosystem businesses, achieving AI handling functions such as ordering food, shopping, and booking flights. This upgrade has launched over 400 AI handling functions and is available for testing by all users. Qwen has improved real-time tool building, multimodal understanding, and ultra-long context processing capabilities, enabling AI to understand interfaces and perform complex multi-step planning. Currently, Qwen has deeply integrated into daily life and travel scenarios, connecting 50 government services including visas and housing provident fund, and can call Fliggy and Gaode to plan itineraries or even make phone calls to book restaurants.

image.png

AiBase Highlights:

✅ Qwen App has fully integrated into Alibaba's core ecosystem businesses, achieving AI handling functions.

🚀 Launched over 400 AI handling functions, users can test directly.

🌐 Deeply integrated into daily life and travel scenarios, connected to government services and supports multi-step planning.

4. Robots can also "take temporary jobs": the robot rental platform Jingtianzu has completed financing, with user count exceeding 200,000

Jingtianzu, a robot rental platform initiated by leading companies in embodied intelligence, has solved the pain points of high costs and fragmented scenarios in the robot industry through an innovative business model. The platform has completed its seed round of funding and plans to expand its service network nationwide to achieve greater market influence.

image.png

AiBase Highlights:

🤖 Jingtianzu uses a "shared rental + platform scheduling" model, reducing the barriers to robot application.

📈 As of now, Jingtianzu has exceeded 200,000 registered users, with daily orders stable at over 200 per day.

🚀 Jingtianzu plans to connect more than 10 body manufacturers by 2026, develop 200 service providers, and cover 200 cities across the country.

5. Video compression rate reaches 0.02%: China Telecom launches generative video compression technology GVC

China Telecom's Artificial Intelligence Research Institute (TeleAI) has released a breakthrough generative video compression technology - GVC (Generative Video Compression), which increases the video data compression rate to 0.02%. The core logic of this technology is "using computation to save bandwidth", by transmitting "instructions on how to draw the image", rather than complete image pixels, thus achieving extreme compression. GVC performs significantly better than traditional algorithms on authoritative datasets. At the same visual quality, traditional methods consume six times more bandwidth than GVC. Currently, the model can achieve near-real-time generation speed on consumer-grade graphics cards and is expected to solve high-definition video transmission issues in extreme network environments in the future.

image.png

AiBase Highlights:

🧠 Shift in logic: Changed the traditional pixel transfer model, by transmitting high-dimensional semantic tokens and using generative AI to "redraw" videos at the terminal.

📉 Extreme compression: The technology compresses video to 0.02%, allowing a 1GB video to be restored on the receiving end with just 200KB of data.

⚓ Wide application scenarios: Designed for extremely low bandwidth environments, applicable to satellite communications, long-distance voyages, and disaster scene rescue in extreme signal conditions.

Details: https://www.arxiv.org/abs/2512.24300

6. Domestic flagship phones collectively bet on AI physical buttons? Honor Magic8 has gone first, with five manufacturers entering an escalated competition

The article discusses the attention paid by domestic smartphone manufacturers to AI physical buttons and the potential changes in interaction methods they may bring.

image.png

AiBase Highlights:

📱 AI physical buttons have become a focus for domestic smartphone manufacturers, possibly redefining the relationship between humans and smart terminals.

🔄 The Honor Magic8 series has introduced a dedicated physical button, supporting quick camera activation with a double click and deeply integrating the self-developed YOYO intelligent entity.

🌐 AI physical buttons are not just a trigger switch but also a "neural synapse" that connects hardware, systems, and multi-device collaboration.

7. ByteDance accelerates hardware layout: Doubao AI earphones produced by Goertek, AI glasses mass-produced 100,000 units, phone assistant prototypes sold out

ByteDance is accelerating its layout of the Doubao AI hardware ecosystem, showing its strategic intent to build a full-scene AI entry point from earphones to glasses and then to phone assistants.

image.png

AiBase Highlights:

📱 The Doubao AI earphones are produced by Goertek, indicating ByteDance's emphasis on product maturity.

👓 The first generation of AI glasses plans to produce 100,000 units, using the Qualcomm AR1 dedicated chip, highlighting the demand for high-performance AI computing power.

📲 The Doubao phone assistant technical preview version is integrated into the Nubia M153 engineering prototype, which was sold out reflecting developers' demand for edge-side AI capabilities.

8. Directly confronting Google Translate: OpenAI secretly launched a dedicated translation website, supporting images and documents

OpenAI launched ChatGPT Translate, a powerful and interactive tool directly targeting Google Translate. This tool supports multimodal input, prioritizes paraphrasing, and allows conversational fine-tuning, showcasing OpenAI's strategic shift in consumer products.

image.png

AiBase Highlights:

🖼️ Supports multimodal input, including text, voice, files, and images

📝 Prioritizes paraphrasing, preserving the deep meaning of the original and supporting tone adjustments

💬 Conversational fine-tuning, allowing users to ask follow-up questions and make modifications to the translated content