AI

Chinese lab targets one of AI’s biggest bottlenecks with bidirectional parallelism and dynamic load balancing. Training large AI models has always been a high-stakes game of computational Tetris—fitting layer after layer of parameters across GPUs without leaving processors idle or networks clogged. This week, Chinese AI lab DeepSeek

IBM has unveiled Granite 3.2, its latest open-source AI model family designed for enterprise use cases—and it’s making bold claims about performance relative to much larger systems like GPT-4o and Claude 3.5 Sonnet. The release introduces reasoning enhancements, multimodal document analysis tools, slimmer safety models, and

One month after releasing its cost-efficient DeepSeek-R1 language model—Chinese AI developer DeepSeek took an unexpected turn: open-sourcing three critical components of its machine learning infrastructure over consecutive days. The latest release targets one of AI’s most fundamental operations—matrix multiplication—with a CUDA-powered library claiming record-breaking performance on

The move could balance server demand while courting global developers. Could it trigger a pricing war in an already cutthroat AI market? DeepSeek, a rising Chinese AI startup challenging Western giants like OpenAI and Meta’s Llama, has unveiled an aggressive discount program offering up to 75% off API access

Despite crowded AI video synthesis field, Alibaba-backed project demonstrates breakthrough efficiency A new contender has entered the increasingly competitive AI video generation space with surprising capabilities that challenge commercial alternatives. Wan2.1, an open-source suite of video foundation models developed by Alibaba researchers now available on GitHub, promises Hollywood-grade output

Anthropic's latest release blurs line between instant answers and methodical problem-solving Anthropic has launched Claude 3.7 Sonnet—a new breed of AI model combining rapid-fire responses with deliberate reasoning capabilities in a single system. The release comes just six months after Claude 3.5 Sonnet redefined expectations

In a move poised to reshape how artificial intelligence systems process complex tasks across distributed networks, Beijing-based AI company DeepSeek has unveiled DeepEP – an open-source library tailored for optimizing Mixture-of-Experts (MoE) architectures. Released during day two of its Open Source Week initiative (GitHub repository), this toolkit addresses one of AI’

Beijing—Just months after making waves with its open-source R1 reasoning model that rivaled OpenAI’s proprietary systems under a fraction of the cost, Chinese AI startup DeepSeek is reportedly accelerating development of its next-generation model codenamed "R2." Originally slated for May 2025, internal sources indicate the company

New optimization techniques approach theoretical hardware limits while slashing memory overhead As tech giants race to optimize increasingly complex AI models for real-world deployment, Chinese AI firm DeepSeek has thrown its hat in the ring with FlashMLA, an open-source decoding kernel promising record-breaking performance on NVIDIA’s Hopper architecture GPUs.