Databricks: shifting just 20% of coding traffic to OSS models dramatically cuts inference spend

Yuchenj_UW · x · 2026-09-19

Databricks researcher Yuchen Li says more engineers there now use OSS models as daily drivers, barely reaching for Claude or Astra. Inference customers see the same: moving even 20% of internal coding traffic from Claude/GPT to OSS dramatically cuts spend. He predicts massive enterprise OSS adoption ahead.

Original post →

More from Models

Models channel →