[bagdaerdev]

Fair point. The first run covers what our customers deploy most via Ollama today, which skews to the Llama / Qwen 2.5 / Mistral / DeepSeek families. Qwen 3.6 and Gemma 4 are top of the queue for the next run same method, same raw JSON. Which quants would be most useful to you: Q4_K_M only, or Q8 as well?

[kennywinker]

I knew there had to be a reason. I guess if you build a functioning system on qwen 3.5, upgrading to 3.6 isn’t necessarily worth the engineering effort.

For me, i don’t personally know anybody with enough vram to be running more than a 4bit quant - so that’s my line.