Anthropic's Claude family of AI models has run into an unexpected snag, according to a report from the tech outlet Let's Data Science, surfaced via Google News. The report says newer Claude models show a "tool-calling regression" — meaning that on this specific capability, recent versions perform worse than earlier ones.
Tool calling is one of the most important skills a modern AI model has. It is what lets a chatbot move beyond conversation and actually do things: run a calculation, search the web, query a database, call an external service, or trigger a step in a larger automated workflow. When a model "calls a tool," it decides which function to use, formats the request correctly, and passes along the right information. Developers building AI assistants and agents lean heavily on this ability, so it is closely watched.
A "regression," in software terms, is when something that used to work well gets worse after an update. According to Let's Data Science, that is the pattern being observed with the newer Claude releases on tool-calling tasks — an area where users would normally expect each generation to improve, not slip.
The source item does not provide the underlying benchmark numbers, the specific model versions affected, or a response from Anthropic, so those details remain unconfirmed here.
Why it matters: as companies race to build AI "agents" that carry out multi-step tasks on their own, reliable tool calling is the foundation the whole system rests on — and a reported step backward in a leading model is a reminder that newer does not always mean better.