Search AI Products and News

To ease cost pressure, OpenAI is testing ads in its chatbot, targeting $2.5B revenue this year and $10B by 2030. However, eMarketer predicts the global standalone chatbot ad market will reach only $5.4B by 2030, leaving a ~90% gap, making the goal unlikely.....

Visual large models faced hidden barriers—SOTA required massive data and compute, reserved for top teams. On July 21, XPeng unveiled TuringViT, an efficient vision encoder that restructures architecture, data, and training, enabling reproducible, low-cost SOTA vision Transformers and democratizing advanced visual capabilities.....

Google DeepMind released GenCeption, a pretrained video generation model that unifies depth estimation, image segmentation, 3D pose estimation, etc. Its performance rivals or surpasses specialized models with much less training data. This challenges the landscape dominated by models like Segment Anything and Depth Anything.....

SenseTime open-sources SenseNova-Vision-7B-MoT, a multi-task vision model integrating object detection, OCR, depth estimation, normal estimation, image segmentation, and multi-view processing in a 7B architecture, providing an efficient base for visual understanding and GUI agents.....

SenseTime has released and open-sourced the "SenseNova-Vision" vision large model, whose core lies in natively integrating vision capabilities into a general foundation model, breaking the fragmented mode of packaging expert models such as detection and segmentation. This model achieves dominant performance across multiple evaluation tasks in four fields, marking a critical upgrade in visual tasks toward a unified native architecture.

Ant Group's Robbyant opensources the LingBot-Vision model family, which achieves outstanding performance in dense space perception tasks through self-supervised vision Transformers and innovative boundary modeling. It surpasses large models with several times more parameters in multiple metrics, breaking the limitations of existing visual foundation models that focus heavily on object recognition, making precise perception of physical space by robots a reality.

On July 7, Ant Group's Lingbo Technology released LingBot-Depth 2.0, a spatial perception model trained on 150M data. It enhances edge clarity, small object recognition, long-range depth estimation, and robustness in complex scenes. Serving as robot eyes, the v1.0 solved transparency and reflection issues; v2.0 further upgrades.....

Cloudflare has released new service regulations requiring all AI vendors to separate search crawlers from training/agent-specific crawlers by September 15th. Mixed crawlers accessing pages with advertisements will be automatically blocked. This rule applies uniformly to new customers, existing users creating new sites, and all free users; website administrators must manually modify backend configurations to allow crawling, directly affecting the standardization of AI crawlers in the industry.

OpenAI hires former Apple hardware executive Paul Mead, who led the Vision Pro team and smart glasses development. He now leads AI device R&D in the hardware division, signaling an accelerated push into AI-native hardware.....

Intensified competition in AI: After Anthropic secretly filed an S-1 draft, OpenAI recently submitted an IPO application to the SEC, moving toward public capital markets. Despite earlier beliefs that being private was advantageous, the company decided to pursue an IPO after weighing the competitive business landscape.....

OpenCV has achieved a milestone upgrade with the official release of OpenCV 5. This version modernizes its overall architecture based on more than two decades of technical accumulation. As a fundamental tool in the fields of computer vision and artificial intelligence, OpenCV is widely used in robotics, embedded vision, industrial inspection, medical imaging, and AR/VR applications. It has received over 86,000 GitHub stars.

Global AI is undergoing an 'AI-native' technological shift. Addressing the current 'language-centric, externally patched vision or speech' architecture, a team released and open-sourced the native multimodal large model LongCat-Next and a discrete tokenizer, aiming to break modal barriers and enable AI to understand the physical world like processing text, achieved by reconstructing the underlying architecture.....

Alibaba launches Qwen3.7-Plus, a next-gen multimodal agent model, accelerating domestic LLM evolution toward embodied intelligence and advanced agents. It inherits Qwen3.7's text capabilities, achieves multimodal breakthroughs, and provides a core foundation for edge and complex workflow applications.....

Aitek has transformed from a traditional hardware supplier into a provider of visual infrastructure in the era of physical AI, focusing on solving the core bottleneck of "enabling robots to understand" the physical world. The company precisely positions itself in the new global technological competition field of physical AI through its full-stack technology matrix, aiming to become an industry intermediary and promote the practical application of large models in real spaces.

Tencent's security team for the shooting game 'Delta Operation' announced a special crackdown on the recently emerged 'AI Vision Adhesion' hacks. These hacks use AI vision recognition technology to capture game footage through streaming software like OBS, identify target positions, and assist in aiming, avoiding traditional methods of modifying memory or files, seriously disrupting high-level competitive balance.

ByteDance open-sources Lance, a native unified multimodal large model with only 3B activated parameters, breaking the technical barriers between understanding models (VLM) and generation models (DiT/Diffusion). It achieves full functionality with extreme lightweight design, challenging the current industry trend of stacking parameters or assembling models, marking an important breakthrough in technological innovation.

Apple launched new accessibility features on the 15th Global Accessibility Awareness Day, integrating Apple Intelligence technology to enhance the experience of people with disabilities on iPhone, Mac, and Vision Pro. The VoiceOver feature has been upgraded, adding an image resource manager, allowing visually impaired users to identify image content through AI, providing more convenient visual assistance.

Apple announced a series of new AI-assisted features driven by Apple Intelligence on May 19th, and introduced revolutionary eye-controlled wheelchair technology through Apple Vision Pro, marking a major breakthrough in the field of accessibility. This update deeply integrates edge AI, upgrading core assistive features while ensuring privacy, such as 'Narration' which uses AI to provide more detailed descriptions of visual content like bills and photos.

The China Mobile Vision Network AI Ecosystem Signing Conference was held in Chongqing with the theme 'Vision Connects Industries, Digital Intelligence Creates the Future,' hosted by China Mobile IoT Company. Experts from six universities and research institutions in Beijing, Tianjin, and Chongqing, along with over 50 partners, attended, marking the launch of deep integration between vision networks and AI. The head of China Mobile IoT's video dep....

Apple Silicon's local AI ecosystem gets a major update: oMLX framework v0.3.9.dev2 integrates Gemma4's MTP visual path, DFlash engine, and ParoQuant quantization, boosting multimodal processing speed and usability, strengthening Apple's edge AI competitiveness.....