GPT-5.6 Sol beats Claude Fable 5 on cost-per-result

GPT-5.6 Sol uses half the tokens of Claude Fable 5 and costs 3x less per output. Token efficiency often matters more than benchmark scores for production workflows.