OpenAI CEO Sam Altman revealed plans to explore a new open-source strategy during a Reddit AMA, sparking industry attention. While not a top priority, it remains key for future development. OpenAI may open-source older AI models, promoting tech adoption but raising business concerns over potential secret leaks. Altman will travel globally to discuss AI trends with stakeholders.

Under competitive pressure, OpenAI accelerates new feature releases. On February 3, 2025, OpenAI introduced "Deep Analysis" for ChatGPT Pro users, generating professional reports via online analysis. This feature acts as a "personal expert advisor," boosting efficiency while reducing errors compared to existing models. Though occasional inaccuracies may occur, OpenAI is optimizing algorithms for better performance. The launch highlights OpenAI's commitment to AI innovation amid increasing compet

Tencent's self-developed DeepSeek-R1 large model now supports one-click deployment on Tencent Cloud's 'HAI' AI platform, enabling developers to access it in three minutes. The 'HAI' platform simplifies complex setup steps, lowering the AI usage threshold and promoting broader adoption. This integration highlights Tencent's commitment to advancing AI technology with efficient and intelligent solutions for developers and enterprises.

Sam Altman, CEO of OpenAI, is embarking on a global tour to engage with investors, developers, and industry leaders in cities like Tokyo, New Delhi, Dubai, and Germany. The trip aims to address rising AI competition, particularly from DeepSeek, a Chinese firm launching cost-effective models. Altman will discuss continuous investment in GPT development and the importance of projects like Stargate, a joint venture with Oracle and SoftBank for advanced data centers. Rumors suggest new financing tal

Amazon Bedrock and SageMaker AI now support DeepSeek's latest R1 model, offering users greater flexibility for diverse AI applications. Bedrock simplifies pre-trained model integration via APIs, while SageMaker enables advanced customization, training, and deployment. Users can also deploy the cost-effective DeepSeek-R1-Distill model using Amazon EC2 and Inferentia/Trainium. Recent DeepSeek models, including R1-Zero (671B parameters) and Janus-Pro-7B for visual tasks, enhance their product line

OpenAI launches the new o3-mini AI inference model for ChatGPT and API services. Free users can access a limited version for the first time, with performance improvements including 24% faster response times. Developers can use it via APIs, while paying customers get additional options like o3-mini-high. Pro users enjoy unlimited access, and message limits for Plus/Team plans are increased to 150 daily. Free users can try the model via the Reason feature.

Microsoft AI CEO Mustafa Suleyman announced on X that all Microsoft Copilot users can now access OpenAI's o1 inference model for free. The Think Deeper feature, previously exclusive to paid users, is now widely available, offering deeper and more accurate responses through multi-step reasoning. This enhancement aims to boost work efficiency and highlights Microsoft's commitment to AI innovation.

DeepSeek, a Chinese AI company, faces scrutiny from OpenAI over allegations of using its data to develop cost-effective AI models. OpenAI and Microsoft are investigating potential misuse of API data, with suspicions of "distillation" technology for training cheaper models. OpenAI accuses DeepSeek of violating terms, despite past controversies over its own data sourcing methods. The issue highlights tensions in global AI competition and intellectual property protection.

DeepSeek launched Janus-Pro, an advanced multimodal model on GitHub, expanding into text-to-image generation. It improves upon its predecessor with optimized training, a larger dataset, and enhanced stability. Janus-Pro outperforms Stable Diffusion and DALL-E 3 in benchmark tests. All Janus models are now open source, promoting innovation in the field.

Apple plans to launch iOS 18.4 in April 2025, featuring significant Siri upgrades including personal context awareness, screen perception, and app access capabilities. The update will expand language support for intelligent features, adding Chinese, Indian English, French, German, and more. Additionally, iOS 18.4 will introduce new emojis such as fingerprints, faces with eye bags, harps, and root vegetables. Details are subject to official confirmation.

xAI's Grok-3, currently in internal testing, is set for release next week. It outperforms OpenAI's o1 and DeepSeek's R1 in reasoning. Designed as a truth-seeking AI, Grok-3 emphasizes factual responses with data and evidence, using the X platform as its primary source. The tech community eagerly awaits its debut.

Gemini 2.0 Flash enhances AI reasoning, math, and multimodal capabilities with 65%-75% accuracy in math tests and an extended context window of up to 1 million tokens. It supports local code execution via API for functional applications but requires occasional debugging. The model excels in ethical reasoning yet needs adaptability improvements. Available for free via API, it offers significant advancements for developers and researchers while addressing data privacy concerns.

US restrictions on Nvidia chips have inadvertently boosted Chinese AI firm DeepSeek, whose R1 model ranks globally among the top 10. Developed using less powerful chips, R1 matches OpenAI's performance at significantly lower costs and is highly popular in open-source platforms. DeepSeek's innovation, driven by resource scarcity, has drawn praise from industry leaders like Microsoft CEO Satya Nadella, highlighting China's growing AI influence and potential to reshape the global tech landscape.

Alibaba launched Qwen Chat v0.2 with web search, video creation, and image generation features. It also released the Qwen2.5-1M open-source models, capable of processing up to one million tokens. An advanced inference framework using sparse attention speeds up processing by 3-7 times. The update enhances multimodal AI tools for professional and creative tasks, with robust multilingual support and extended context lengths. Users can access these via Alibaba Cloud platforms.

Eleven Labs, a synthetic voice tech startup, secured $250M in Series C funding, valuing it at $3B-$3.3B. Led by ICONIQ Growth, the round included Andreessen Horowitz. The company offers a cloud-based AI platform generating over 1,000 synthetic voices for dubbing, audiobooks, and narration, with voice cloning capabilities in 20+ languages. Despite praise for its tech, Eleven Labs faces criticism after recreating the voice of late French actor Alain Dorval without family approval. Clients include

The Davos Forum highlighted six key AI topics, including intensified global AI competition, AGI advancements, enterprise AI applications, new architecture paradigms, AI in drug development, and infrastructure investment. Industry leaders predict transformative breakthroughs within 2-5 years, emphasizing AI's role in reshaping industries and boosting national competitiveness. Discussions reflect both optimism and caution as AI moves from experimentation to a powerful driver of societal change.

ByteDance launches UI-TARS, a next-gen GUI agent model for natural language interaction with desktops, mobile, and web. It features real-time perception, reasoning, actions, memory, cross-platform compatibility, visual recognition, and automated task execution. Supports cloud and local deployment, trained on large-scale datasets for precise interface understanding and multi-step reasoning. Expected to enhance user productivity significantly.

Zhipu's GLM-PC, based on CogAgent, mimics human capabilities for efficient computer tasks. It integrates code generation and GUI comprehension, excelling in task planning, execution, and self-correction across platforms. Supporting multimodal processing, it enhances information management and offers seamless user experiences in various applications.

Baichuan Intelligence launches Baichuan-M1-preview and Baichuan-M1-14B AI models excelling in medical and deep reasoning. Baichuan-M1-preview offers superior math, coding, and medical reasoning; Baichuan-M1-14B is an open-source medical-enhanced model outperforming larger counterparts in clinical evaluations enhancing AI medical ecosystem.

OpenAI upgrades Canvas with o1 model, HTML and React support enhancing code efficiency. Now on macOS, offering better local environment privacy and faster response for developers.