Sort:  

My understanding it is Llama. It looks like it is about 2 years old on the data used for training.

Thank you for the information. DeepSeek came much later to the scene. Llama 3.X is the most likely case.