Same issue here. I have input costs of 1-2$ per prompt, because it loads tons of files unrelated to the prompt. Before, I could simply use sm model, which would efficiently just be around 0,10$ per prompt. At this point, v0 is getting indrecibly expensive!!! I was super hyped for the integration of gpt-5, which was sadly replaced alltogether by the Agent. I don’t see any use in the agent.
Many are experiencing this issue and have reported it widely in these forums. I’m sure the v0 team will have seen our cries for help and will be discussing a fix.
It seems that at this point we can only wait. Even the prompt provided by the v0 staff didn’t work for me. I’ve already reported the issue, but there’s nothing more I can do. All I can do now is hope that the v0 team fixes it—or consider switching to another platform.