What are the costs to get equivalent work done compared to the Claude code max 20x sub?

I would say GPT-5.6 Sol is miles better at least in areas like legal work. Fable is still making stupid mistakes and display no understanding of what it reads and writes. Fable often makes logic and grammar mistakes that change the meaning of sentences, so you have to read everything carefully.