Qwen3.8-flash-next can run in ~60gb, offloading 50gb to ssd, and is comparible.

The business has to downgrade to be sustainable, and open models prove it.