Found a new local model that's supposedly 10x faster at reading executive sized documents and also has big brain for long conversations and is smarter than my regular model.
It's qwen 3.8 flash next iq3_s