This model is a 8-bit quantized version of the MNN model exported from OpenELM-3B-Instruct using llm-export.