When using the GPT model series, turning on max mode does not increase the context window. So does that mean enabling max mode for GPT models only raises the cost by 20% without improving any other performance? Then isn’t this button pretty much a rip-off for the GPT model series?