DSPy Introduces Flex: Code Optimization Boosts Accuracy, Cuts LLM Calls by 75%

lateinteraction · x · 2026-08-06

The DSPy framework introduced Flex, a new feature that optimizes program code alongside prompts. In a cited example, the GEPA optimizer improved task accuracy from 90% to 95% while reducing LLM calls by 75%. This self-optimizing harness significantly enhances accuracy and slashes inference costs, bringing the expense of processing 1,000 records down from $0.98 to just $0.01.

Related event: DSPy Introduces Flex for Code-Level Optimization, Slashing LLM Calls(6 posts)→

Original post →

More from coding & agent

coding & agent channel →