Google's Gmail Live, announced at I/O, is now in beta testing, allowing users to search their email inboxes using voice commands. This Gemini-powered feature will initially be exclusive to Google AI Pro and Ultra subscribers, replicating Gemini Live's voice chat experience in apps like Gmail, Docs, and Drive.
DeepSeek has begun limited beta testing of its new image recognition function, announced by multimodal team leader Chen Xiaokang on Wednesday. The feature is available on the company's chatbot website and mobile app, adding to existing "expert" and "flash" chat modes introduced earlier this month. The release comes days after DeepSeek unveiled its V4 flagship model followed by extensive price cuts. The Hangzhou-based startup, which gained international recognition in January 2025 for its cost-effective reasoning model, has faced criticism for lacking multimodal capabilities despite its rapid growth.
Alibaba's new AI video generation model, HappyHorse 1.0, has surpassed ByteDance's Seedance 2.0 in benchmark rankings for both text-to-video and image-to-video capabilities, highlighting the intense competition in the AI talent race.
