Tested different models with Clawberto this week. Currently finding MiniMax M2.5 to be the most cost efficient for general agent tasks. Claude models are the only ones giving semi decent results for his @farcaster posts. With weaker models he gets confused way too much.