Benchmarking LLMs Can Reduce API Costs by 80% or More

We benchmarked 100+ models on our actual task and found a much cheaper alternative that works just as well.