“Public LLM benchmarks are not an accurate reflection of a model's utility value for enterprise customers.”