DeepSeek Pitches a New Path to Scaling AI, but Researchers Urge Caution

A fresh proposal in the race to scale artificial intelligence DeepSeek has unveiled a new approach to training large artificial intelligence models, presenting what it describes as a potential alternative to the parameter heavy scaling strategies that dominate the industry today. The proposed

DeepSeek Opens 2026 With Ambitious Plan to Train Larger AI Models at Lower Cost

A strategic start to the new year Chinese artificial intelligence start up DeepSeek has begun 2026 by releasing a technical research paper that signals a clear strategic direction for its future development. Co authored by founder and chief executive Liang Wenfeng, the paper