Google Research has released TimesFM-3, a 330-million-parameter forecasting model that analyzes time series data alongside related external variables such as weather forecasts and scheduled sales promotions (The Decoder). Rather than predicting future values step by step, the model fills all future time points in a single forward pass, addressing a common use case in demand planning and inventory forecasting where promotional events and seasonal factors are known in advance.
Anthropic has published a new plugin evaluation workflow for Claude Code that includes six grader types, a no-plugin baseline for comparison, and a continuous integration gate for skill assessment (MarkTechPost). The claude plugin eval command runs plugins against realistic prompts, grades outputs, and measures performance relative to unaugmented Claude runs. In related work, Cohere released North Small Translate, an open-weight 218-billion-parameter Mixture-of-Experts translation model covering 50 languages with a score of 83.6 on Cohere’s WMT26 evaluation and weights available for non-commercial use (MarkTechPost).