I think the cost calculation is based on the official token pricing. But the actual cost is coding plan. Both GPT and Kimi coding plan are generous. The the task performance is the first priority, token cost within coding plan is secondary
yes! but we are continuing to optimize the token efficiency. and I think open source model token will get cheaper and stronger. so eventually will bring the cost down!