The Inference Shift

If it's that much faster then perhaps it can come up with 10 solutions and rank them and still have a total better performance than a larger LLM.