Google AI Studio launches Gemini-2.0-flash-live-001 real-time multimodal model, supporting streaming audio and video processing, helping to develop low-latency AI applications.
Cogito v1 Preview is fully open source, covering 3B to 70B models, supports function calls and inference patterns, and has a comprehensive leading performance.
Zhixiang Future has released HiDream-I1, a 17B parameter open source image generation model, with excellent image quality and first-class prompt word compliance capabilities.
MagicColor launched by the Hong Kong University of Science and Technology uses diffusion models and self-supervised training to achieve efficient multi-instance line-brush coloring, which is suitable for animation production, digital art, game development
Chinese company DeepSeek launches the high-performance open source AI model R1, which enables powerful inference capabilities at a cost far lower than OpenAI.
Sync Labs launches Lipsync-2, the world's first mouth-synced AI model that does not require training, suitable for live-action, animation and AI content generation.