AI Signal 333
llm-anthropic 0.26
↗
Illustration only Photo by Magnus Engø on Unsplash
The llm-anthropic plugin updated to 0.26, adding three new Claude 5 models and server-side tool support while simplifying the thinking/reasoning interface and removing several older options.
The server-side tools (WebSearch, WebFetch, CodeExecution, AnthropicMCP) let you invoke Anthropic-hosted capabilities directly through the LLM CLI or Python API, removing the need to wire up your own tool implementations for those functions. The thinking model changes mean existing scripts using the old thinking_budget, thinking_display, or thinking_adaptive options will break and need migration to the simplified thinking and thinking_effort parameters.
Written by elseif from the cluster below · every claim links back to a sourceThe three things worth knowing
Three new Claude 5 models are now available: claude-fable-5, claude-sonnet-5, and claude-opus-5, with thinking enabled by default on all of them.
Server-side tools (WebSearch, WebFetch, CodeExecution, AnthropicMCP) replace the previous -o web_search* options and are accessed via the -T flag or Python tools= parameter.
The extended thinking interface was simplified to two options (thinking and thinking_effort), and the older thinking_budget, thinking_display, and thinking_adaptive options were removed.
THE CLUSTER