The a16z Show · Friday, July 31, 2026
Jesse Zheng of Decagon argues that fine-tuned open-source models can surpass even state-of-the-art closed-source models in specific tasks, offering advantages in performance, cost, and speed. He emphasizes that for specialized enterprise needs, smaller, tailored models are often more effective than large, general-purpose ones.
“So when we find tune smaller, dumber models, it's that they start as general purpose. But on the specific task we want them to do, they actually outperform the large, smart state of the art model.”
“So we end up getting all three things. It is better at the task, it is cheaper, and it is faster.”