Yes, reasoning models can potentially be optimized with our uncertainty estimations. We are currently testing the library with DeepSeek R1