qwen2 7b

7B instruction-tuned LLM by Alibaba, the Qwen2 generation — since superseded by Qwen2.5 and Qwen3, both also in this catalogue. It is here mainly as a comparison point: one recipe, on an 8 GB RTX 3060 Ti, where a shared 4-bit Ollama run records 63.73 tokens/s, third of the nine models measured on that same card. Apache-2.0.