All AI News

Stepfun AI's Step-3.5-Flash model has created a shockwave in the AI world by delivering performance far beyond its size, reportedly surpassing even 120B parameter giant models. Users note this 'monster' can compete with rivals like DeepSeek-V3.2 and is available with free access.

Moltbook, a social network designed specifically for AI agents, created a brief phenomenon in the tech world. This interest stems from dynamics of 'ownership' and display reminiscent of the 1990s Pokémon collection craze. Users participate in a social arena by sharing their own created intelligent agents on this platform.

While OpenAI celebrates ChatGPT's growth, users are turning to price-performance-focused alternatives. ZDNET and Android Authority analyses reveal Google Gemini's rise and intensifying competition in the AI assistant market.

Western tech giants' focus on closed-source models has created a leadership opportunity for Chinese-origin models in the open-source AI arena. Strengthened by strategic state policies, academic collaborations, and pragmatic approaches, Chinese models are gaining significant advantages in performance and accessibility.

A developer has created a local home automation assistant called 'Fulloch' based on Qwen3 that processes all operations on the user's own computer. The project eliminates cloud dependency, promising superior privacy and speed.

Teradata, a leader in data analytics and cloud solutions, has brought its enterprise AI agents to Google Cloud Marketplace. This strategic move will enable businesses to conduct advanced AI operations through a secure and scalable cloud infrastructure. The collaboration opens the door to a new era of hybrid cloud and AI integration.

Microsoft researchers have proposed the 'OrbitalBrain' system, which enables running distributed machine learning on satellite constellations. This system aims to establish an artificial intelligence network that continuously learns in space and operates through inter-satellite communication. The project has the potential to revolutionize data processing and analysis.

OpenAI has begun testing sponsored responses in ChatGPT's free version. Links labeled 'sponsored' won't influence the AI's answers but open doors to a new revenue model for sustaining the free service. This marks a significant shift in the platform's business approach.

Microsoft and Tsinghua University researchers have developed a new training technique called 'Pull-Push' that significantly reduces 'hallucination' errors made by AI models in visual interpretation. This method enhances AI reliability by enabling models to understand images both holistically and in detail. The development is considered a major advancement in computer vision and visual-AI fields.

Anthropic's Claude Code platform is transforming software development processes with autonomous coding agents called 'Subagents'. This system initiates a new era in human-machine collaboration by dividing a single task among AI agents with different specializations.

China's humanoid robot 'Bolt' has set a world record by reaching a speed of 10 meters per second, officially becoming the first robot to surpass human biological limits. This milestone achievement is being hailed as the beginning of a new era in robotics technology.

Researchers are combining the next-generation generative AI technique called Flow Matching with a 'topological' perspective to understand complex structures like brain graphs. This approach holds potential not only for visual arts but also for creating data that preserves structural integrity, from scientific data to new molecules and materials.

A new era of competition has begun in Silicon Valley for artificial intelligence researchers. Talented scientists are switching companies with million-dollar transfer fees, similar to professional athletes. This trend is fundamentally transforming the tech industry's work culture and human resource strategies.

OpenAI has launched advertising tests for ChatGPT's free and Go plans in the United States. These ads, displayed based on user queries, offer clues about the platform's future business model while raising questions about the sustainability of free access. The company pledges that advertisements will not compromise the independence of responses.

A multimodal LLMOPS project developed on Microsoft Azure using LangGraph and Azure AI services is revolutionizing video content moderation. The system employs OpenAI's GPT-4o model as a 'reasoning engine' to automatically detect compliance violations and brings transparency to the entire process. The project aims to minimize human errors in content moderation while increasing efficiency.

The addition of GLM 5 support to Hugging Face's Transformers library signals that Zhipu AI's new model launch is approaching. Industry excitement is building over claims that OpenRouter's 'Pony Alpha' model could be a covert GLM 5 distribution, highlighting the growing global influence of Chinese-origin large language models.

Comprehensive tests comparing leading XR glasses models from Viture, RayNeo, and Xreal have revealed a surprising industry result: the product with the highest price tag failed to deliver the best user experience. The research showed that beyond brightness and technical promises, factors like comfort, software compatibility, and daily usability are decisive.

Anthropic's Claude Code is ushering in a new era in AI-assisted coding. Experts indicate this system, ranging from slash commands to intelligent agents, will fundamentally transform software development processes. The tool aims to boost developer productivity through codebase comprehension, task automation, and Git workflow management.

The common English word 'more' functions as far more than a simple quantifier in constructions like 'more than one' or 'more likely than not.' Linguistic analysis reveals these phrases carry deep layers of meaning and cultural coding. This article explores the subtleties of this linguistic metaphor and the role of translation tools in conveying these nuances.

A new study reveals that even the most advanced multimodal AI models like GPT-4o and Gemini struggle to exceed a 50% success rate in basic visual recognition tests. This finding highlights the critical gap between AI systems' claims of 'seeing' and their actual comprehension abilities.

The rapid rise of artificial intelligence in visual production is prompting questions about the human role in 3D art and modeling. Experts emphasize that technology is not a threat but a tool that complements human creativity, noting that creating 'intent' and 'context' remains uniquely human.

ByteDance has announced Protenix-v1, an open-source AI model capable of protein prediction at the level of AlphaFold 3. This development could usher in a new era in drug discovery and understanding disease mechanisms by democratizing access to biotechnology research.

Experts highlight that excessive anxiety and panic about screen time in digital-age parenting leaves parents with unnecessary guilt. The focus should shift from rigid time limits to the quality and context of consumed content.

Scientists have developed a revolutionary open-source framework to prevent large vision-language models from producing incorrect or inappropriate outputs. The system called 'STLE' teaches artificial intelligence to measure its uncertainties and say 'I don't know' when it lacks confidence. This development is considered a significant milestone for AI reliability and ethical use.