It's commercially more interesting to offer a GPT fintuned in 3.5 than a GPT not fituned in 4. Plus the fees are 4 times lower so I don't think it's a real problem.

We're looking into this but the current issue here is that OpenAI currently only allows fine tunning for gpt-3 and not gpt-4 so thats the main blocker as most people will want to fine tune ontop of gpt4. Actually, this isn't a blocker as I previously thought. We'll just allow fine tuning on top of gpt-3.5