A mid-sized Chinese-developed model that trades some precision for dramatically reduced memory footprint through 4-bit integer quantization. It converses fluently in both English and Chinese, handling everyday reasoning and instruction-following competently, though the aggressive quantization means it occasionally loses nuance on complex multi-step problems compared to its full-precision sibling. Think of it as the pragmatic colleague who fits into tight spaces without too much fuss.