Reasoning | NVIDIA Dynamo Documentation

Some models emit reasoning or thinking content separately from their final response. Dynamo can split that output into reasoning_content and normal assistant content by configuring a reasoning parser. As with tool calling, there are two ways to do this to ensure wide coverage and day0 model support.

Choose a parsing path

Path	When to use	Page
Dynamo	Dynamo ships a Rust parser for the model’s reasoning format. Lowest latency, the default path.	Reasoning Parsing (Dynamo)
Engine Fallback	Use the framework’s implementation (vLLM or SGLang) for pre/post processing, including tool call and reasoning parsing - ensure consistency with framework behavior.	Reasoning Parsing (Engine Fallback)

Start with the Dynamo path. Fall back to the engine path only when Dynamo’s registry does not list a parser for your model.