This model is a 4-bit quantized version of the MNN model exported from QwQ-32B-Preview using llm-export.