Comment by pontussw

It works with all models, some have a cost multiplier like Opus 4.6 ”charges” 3 requests per prompt, but its still only for the prompts you send yourself - even if it works on the issue for hours. GPT-5.4 has no multiplier i.e. costs 0.04$ per prompt.

Worth noting however that they are starting to introduce rate limits lately so you might struggle to run multiple concurrent sessions, though this is very inconsistent for me. Some days I can run 3-4 sessions concurrently all day, other times I get rate limited if I run one non-stop..