Probably even higher because openai and anthropic undoubtably have the lowest cost per token generated, especially with cerebras being able to serve a million tokens every 16 minutes.