DeepSeek-R1-Lite, a homegrown inference model comparable to o1-preview, is online!

Google is testing a new feature in its Search Labs platform called Web Guide. The core idea behind the experiment is to use artificial intelligence to take web links on search results pages and reorganize them in a smarter, more structured way, aiming to simplify the process of discovering information and exploring web pages for users...

SmartScope AI recently released GLM-4.5, its next-generation flagship foundation model designed for intelligent body (Agent) applications and open-sourced simultaneously on the Hugging Face and ModelScope platforms, with model weights following the MIT license. The model family utilizes the Mixed Expertise (MoE)...

In July 2025, the World Artificial Intelligence Conference (WAIC 2025) opened in Shanghai. This year's conference sends a clear signal: the industry focus has shifted from a race of technical parameters to a profound exploration of the role of AI in society. The core topic is no longer what AI can do, but what we want it to be,...

On July 25th, AI video company Runway released its latest research project, Runway Aleph, which differs from its predecessor and competitors' general focus on "text-to-video" generation capabilities. Aleph is defined as a "context-aware video model" (in-context...

As large-scale language modeling (LLM) gradually infiltrates all aspects of academic research, a new kind of academic integrity issue has surfaced. Recently, ICML 2025 (International Conference on Machine Learning), a top artificial intelligence conference, released an announcement that explicitly refers...

With cross-language communication becoming the core demand of globalization, simultaneous interpretation has always been the most challenging peak in the field of machine translation. Recently, Byte Jump Seed team released an end-to-end simultaneous interpretation model called Seed LiveInterpret 2.0, which provides a reliable technology for real-time cross-language communication...

Google has announced the introduction of a series of AI-powered creative tools for its app Google Photos, designed to turn users' static photo libraries into dynamic creative canvases. The core features of the update include "Photo to video" and "Remix," as well as a new "Create" tab...

ByteDance recently pushed its AI application development platform, Coze Studio, to the open source community, a move that is not mere technology sharing but a key strategic layout in the increasingly fierce race for the AI development toolchain. The current market has shifted from the competition of the model itself to the application development platform and ecological...