This model is a 4-bit quantized version of the MNN model exported from SmolLM2-360M-Instruct using llm-export.