This is likely an OpenAI model as it's relatively easy to get the model to admit to it without leading, however it generates tokens at ~124 tokens/s which is far faster than a GPT-5 class model (GPT-4.1 is 93 tokens/s) and its context window is only 256k so it's likely something more unusual, perhaps the open source model.