LLMs can regularly produce correct conclusions from the input they’re given. When a human does that, we call the process involved “reasoning”. LLMs can even explain how their conclusions are supported in ways that again, if a human employed them, would be called reasoning.

As such, functionally speaking - focusing on inputs and outputs - models can be said to reason. If you’re tempted to argue that it’s “only statistics” or similar, consider that human reasoning is only chemical and electrical signals in neurons.

I think I've read some study that the LLM reasoning is a mirage.