Databricks swaps to Chinese AI model GLM 5.2 for coding

Databricks ran a real-world test on its own massive codebase and found that the Chinese open-source model GLM 5.2 performs just as well as Anthropic’s high-end Opus 4.8 for coding tasks—but at a much lower cost. At $1.28 per task versus $1.94 for Opus, the savings are hard to ignore. The company is now making GLM 5.2 the default coding engine for its developers, marking a major shift toward cost-effective open-source models. This isn’t an isolated move: Coinbase slashed AI spending in half by switching to Chinese models, Lindy saved millions by ditching Claude for Deepseek, and Snowflake found similar cost advantages. Chinese models now account for over 30% of weekly traffic on OpenRouter, up from 11% last year, at 60-90% lower cost. Databricks also found that no single model dominates across all tasks—performance tiers vary, and mixing models from different providers actually gives the best bang for your buck. The key insight? Token price isn’t everything—efficiency per task matters more. Why it matters: This is a wake-up call that the AI coding market is no longer a Western monopoly, and open-source models are poised to democratize access to top-tier coding assistance.