In a significant development for the AI coding landscape, Databricks has announced that it will adopt the Chinese open-source model GLM 5.2 as its default coding engine. This move comes after the company conducted an internal benchmarking process on its multi-million-line codebase, where GLM 5.2 matched the performance of Anthropic’s Opus 4.8 at a substantially lower cost — $1.28 per task compared to $1.94.
Performance and Cost Efficiency
The benchmarking results indicate that GLM 5.2 not only matched the high-performance capabilities of proprietary models like Opus but also delivered these results at a fraction of the cost. This cost advantage, combined with the model’s strong performance, makes it an attractive option for Databricks' daily coding operations. The company is now integrating GLM 5.2 into its core workflows, signaling a shift in how enterprises approach AI-powered development tools.
Implications for the AI Industry
Databricks' decision underscores a broader trend in the AI industry: the growing influence of open-source models and the diminishing dominance of a single provider. The company emphasized that relying on public benchmarks can be misleading and encouraged businesses to develop their own internal testing environments to accurately assess model performance. This sentiment reflects increasing industry awareness of the need for customized, reliable evaluation methods in a rapidly evolving AI landscape.
As more organizations turn to open-source models for cost-effective and efficient AI solutions, Databricks’ adoption of GLM 5.2 may serve as a catalyst for further exploration and integration of non-Western AI models in enterprise settings.
