Comment by nsingh2

Also with Qwen 3.8 being more token hungry than Luna, using around 2.3x tokens. Which hurts for local deployment.