Qwen1.5 7B

A compact multilingual workhorse with a particular strength in Chinese and English tasks. At 7.7B parameters it handles everyday reasoning and text generation competently, though it can struggle with complex multi-step problems where larger models have more headroom. It tends to be straightforward and cooperative, making it a practical fit for constrained environments.