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

GPT-5.6 Sol uses half the tokens of Claude Fable 5 and costs 3x less. The article argues token efficiency often matters more than benchmark scores for real workflows.