Are there better agent patterns for powerful reasoning models?

[gaochenxi]

With the improved reasoning capabilities of recent LLMs (GPT-5, Claude opus 4.1, etc.), I'm wondering if the traditional plan-and-execute(https://github.com/langchain-ai/langgraph/blob/main/docs/docs/tutorials/plan-and-execute/plan-and-execute.ipynb) pattern in agent workflows is still optimal.
Key observations:

1. Modern models seem to have implicit planning capabilities embedded in their reasoning process
2. They can handle multi-step tasks with better context understanding in single calls
3. The explicit "planning" step might be redundant when the model already does this internally

With models that have powerful reasoning capabilities, are there already more suitable agent patterns available now?