ChatGPT Teams account. AFAIK, CGPT-4 output is 4k tokens. Playground “latest CGPT-4 model” shows 16k output tokens. Despite all attempts, CGPT limits outputs around 1,300-1,600 tokens.
Does it get exponentially more expensive for them to generate longer responses? Or all the same 10x 1,000 token responses and 1x 10,000 token response?
It is fine with generating endless 1,300 little “response-lets” but refuses to even go up to 4k, let alone beyond.