DATABRICKS PICKS CHINESE OPEN-SOURCE MODEL AS CODING ENGINE

In internal testing on its multi-million-line codebase, Databricks found GLM 5.2 delivered equivalent performance to Opus 4.8 while reducing per-task expenses from $1.94 to $1.28. The company plans to deploy the model as its primary daily coding tool. The decision reflects a broader shift in enterprise AI procurement. Databricks argues that no single provider currently dominates across all use cases, and companies should develop custom benchmarks using their own codebases rather than relying on public leaderboards. GLM 5.2, developed by Chinese AI firm Zhipu AI, represents growing competition in the coding AI space beyond dominant Western providers. The cost advantage and performance parity suggest enterprises have viable alternatives to expensive commercial models for specialized tasks. Databricks' endorsement carries weight in developer circles and may accelerate GLM adoption among companies seeking cost-effective coding solutions.