Gemma 4 is much less popular on Hugging Face than Qwen 3.x.

You're not looking at the right Gemma models, the ones you have listed are the base. You'll need to look at the "it", which are the Instruct models. [https://huggingface.co/google/gemma-4-31B-it](https://huggingface.co/google/gemma-4-31B-it) = 4,465,127 downloads and 110 finetunes [https://huggingface.co/google/gemma-4-26B-A4B-it](https://huggingface.co/google/gemma-4-26B-A4B-it) = 3,285,842 downloads and 54 finetunes. This doesn't even include the GGUF'ed versions which most of the "GPU poorts" are running. You'd want to look at Unsloth (https://huggingface.co/collections/unsloth/gemma-4) or Bartowski (sorry B' doesn't have a collection, so go check https://huggingface.co/bartowski/models). There are others but these guys are the bigger GGUF makers. Qwen has their base and instruct models labeled differently (https://huggingface.co/Qwen/Qwen3.5-35B-A3B-Base vs https://huggingface.co/Qwen/Qwen3.5-35B-A3B). Again, these are not GGUFs, look to Unsloth - https://huggingface.co/collections/unsloth/qwen35 From what I can see, it's actually the opposite, Gemma 4 has more downloads.