LangChain Core 1.5.0 adds standard reasoning_effort parameter

Full analysis

LangChain Core 1.5.0 introduces reasoning_effort as a standardized parameter across chat models. This allows agent builders using LangChain to control the reasoning depth of underlying models (such as OpenAI o-series or Anthropic's extended thinking models) through a single consistent interface, rather than passing provider-specific kwargs. The remaining changes are routine dependency bumps (soupsieve, mistune) with no agent-relevant impact.