DeepSeek-V3.5: 671B MoE Model Surpasses GPT-5.2 on Chinese & English Long-Context Benchmarks
DeepSeek open-sourced V3.5 (671B MoE), setting new state-of-the-art on 1M+ token long-context Chinese and English benchmarks, with native tool-calling and improved multilingual reasoning, making it one of the strongest open-weight models for enterprise long-document processing.
new
Anthropic Claude Full Family Launches on Microsoft Azure Powered by NVIDIA GB300 Blackwell Ultra
Google Gemini Omni Flash Opens Developer Public Beta: Text-to-Video with Native Audio
South Korea Launches Trillion-Dollar AI Chip Initiative Led by Samsung and SK
Meituan Unveils LongCat-2.0: World's First Trillion-Parameter Model Trained Entirely on Domestic Computing Power
OpenAI Unveils GPT-5.6 Series, Outperforming Claude Mythos 5
Meta Launches $299 Own-Brand AI Glasses to Accelerate Consumer AI Device Adoption
NVIDIA Launches BioNeMo Agent Toolkit: AI Deepens Integration into Life Sciences
Anthropic Launches Claude Tag: AI Evolves from Personal Assistant to Team Collaborator
Volcano Engine Launches Doubao LLM 2.1 Pro with Enhanced Multimodal Capabilities
China's "Lingsheng" Supercomputer Tops Global TOP500 List