SiliconFlow secures近2.9 billion RMB funding to scale its self-developed inference engine and multi-modal API platform.
Aggregates penalty announcements from five platforms: the CSRC, Shanghai Stock Exchange, Shenzhen Stock Exchange, National Financial Regulatory Administration, and AMAC. This week features a total of 36 items: 29 from the NFRA and 7 from the SSE.
Through foundational motion models and automated platform, QiaoJie ShuWu enables cross-platform robot dance deployment in 30 minutes, aiming to become infrastructure-level robot OS.
OceanBase's Scout, built on GLM-5.2, achieves 90.62% accuracy, ranking first on the DAB leaderboard.
NVIDIA CEO says paying billions in taxes would be an honor and responsibility.
A full inventory of every video, image, and TTS model actually callable on my production line: 15 DashScope keys tested individually against 7 video endpoints each; on the video line, Wan3.0-video at Arena 1481 (Elo #3) outperforms the pipeline's original Wan2.7 (+53 points); on the image line, GPT-Image-2.5 Sunburst 1421 tops the Arena text-to-image chart—coincidentally the current default; on the TTS line, recommend switching from CosyVoice to Qwen3-TTS-Flash (97ms first packet, voice cloning/design, 9+ dialects). Data cross-referenced across LMArena, Arena.ai weekly, Artificial Analysis, and SuperCLUE; single-engine sources are all flagged with a discount.
Microsoft's Mustafa Suleyman discusses balancing AI regulation with maintaining US technological leadership
Suleyman asserts regulation enables safety and should not be abandoned due to geopolitical competition concerns.
Microsoft engineers use AI agents to port Copilot runtime from TypeScript to Rust for $120K in tokens, achieving 15.9x performance gain
Model Context Protocol serves as an open standard enabling AI agents to seamlessly connect with tools and data sources across ecosystems.
