Description
Are you paying too much for LLMs? In this talk, Yuchen Jin (AI Inference Lead at Databricks) breaks down why the intelligence gap between closed-source frontier models and open-source AI is rapidly closing—and why relying on public benchmarks like SWE-bench or static leaderboards can mislead enterprise AI strategies.
Discover how top engineering teams evaluate large language models (LLMs) using Pareto frontier curves (accuracy vs. cost per task), optimize AI inference performance, and build custom evaluation frameworks (Evals) to drive lower costs and higher speed.
Explore more at: https://www.databricks.com/product/artificial-intelligence/unity-gateway