Tencent's WeChat Vision Team releases WeMM-Embedding, a family of multimodal AI models in 2B, 4B, and 9B sizes that claim top scores on MMEB-v2 and MMEB-v3 benchmarks, supporting text, images, videos, and documents, now publicly available on Hugging Face under Apache License 2.0.
Palona AI secures $20 million in Series A funding to scale its vision AI platform that automates real-time operations for restaurants, hotels, and malls, with early results showing a 20% revenue boost at a restaurant chain deployment.
Berkeley researchers unveil PixelRAG, a groundbreaking open-source AI system that searches documents as images rather than text, preserving critical visual elements like tables and charts across 8.28 million Wikipedia pages using a fine-tuned vision model and publicly available tools on Hugging Face.
A new open-source C++17 port of ByteDance's Depth Anything 3 model runs up to 1.31x faster than PyTorch on CPU, uses half the memory, loads 6.7x faster, and requires no Python, PyTorch, or CUDA — delivering monocular depth estimation and 3D point cloud export from a single image in a compact 99 MB quantized package.
OpenAI's CLIP neural network matches ResNet50's ImageNet performance with zero labeled training data, marking a major breakthrough in computer vision by using natural language instructions instead of task-specific optimization.
AllenAI's open-source olmOCR toolkit converts PDFs and images into clean Markdown text — including equations, tables, and handwriting — for under $200 per million pages, with its latest v0.4.0 release scoring 82.4 on a 7,000+ test benchmark, rivaling top OCR tools while supporting GPU inference, Docker, and multi-node cloud processing.
Moondream's new Photon inference engine eliminates GPU idle time through pipelined decoding, boosting vision-language model throughput by up to 35% on NVIDIA B200 hardware and achieving near-realtime inference at approximately 33ms.
Baidu's newly released Unlimited-OCR, an open-source long-horizon document parsing model built on DeepSeek-OCR, is rapidly gaining traction with 5.6K GitHub stars, offering powerful single-image and multi-page PDF parsing capabilities with a live Hugging Face demo available now.
TwelveLabs launches Rodeo, a groundbreaking AI video editor that condenses 40 hours of footage into a polished reel in minutes using natural language prompts, positioning the tool as a weapon against the rising flood of low-quality AI-generated content online.
A new standalone C++ engine called SAM3DBody-cpp now delivers real-time 3D full-body motion capture from a single camera with zero Python dependencies, supporting multi-person tracking, a 70-joint skeleton, and direct BVH export compatible with Blender.