Databricks has announced that it will make China’s open-source AI model GLM 5.2 its default coding engine for internal software development after the model delivered performance comparable to Anthropic’s Claude Opus 4.8 while significantly reducing inference costs. The move marks one of the strongest endorsements yet of a Chinese open-source AI model by a leading U.S. enterprise AI company and highlights the growing shift toward lower-cost, high-performance alternatives in the rapidly evolving AI market.
According to Databricks, internal testing on its multi-million-line production codebase showed that GLM 5.2 performed in the same top performance tier as Claude Opus 4.8 and OpenAI’s GPT-5.5 for software engineering tasks. However, GLM 5.2 completed coding tasks at a substantially lower cost, making it the company’s preferred day-to-day model for developers.
GLM 5.2 Matches Frontier Models at Lower Cost
Instead of relying solely on public AI benchmarks, Databricks created its own evaluation framework using real engineering tasks from its production software.
The results showed that GLM 5.2 consistently ranked among the highest-performing models for coding while offering a significantly better cost-to-performance ratio.
| Model | Performance | Cost Per Task |
|---|---|---|
| GLM 5.2 | Top performance tier | $1.28 |
| Claude Opus 4.8 | Top performance tier | $1.94 |
| GPT-5.5 (selected configurations) | Top performance tier | Competitive depending on configuration |
According to Databricks, the evidence now supports using GLM 5.2 as the default “daily driver” for coding tasks across its engineering teams.
Why Databricks Made the Switch
The company said public coding benchmarks often fail to accurately represent real-world enterprise software development because models may already have seen benchmark datasets during training.
To address this limitation, Databricks benchmarked the models against its own proprietary codebase, measuring how effectively each model handled practical software engineering challenges rather than synthetic programming tests.
Developer feedback from internal pilots also favored GLM 5.2, with engineers reporting strong coding quality alongside meaningful reductions in AI operating costs.
Open-Source AI Gains Momentum
GLM 5.2 was developed by Chinese AI company Z.ai and has quickly emerged as one of the strongest open-source coding models available.
The model features:
- Strong software engineering performance.
- Up to a 1 million-token context window.
- Open-weight licensing for developers.
- Competitive agentic AI capabilities.
- Lower inference costs than many proprietary rivals.
Its release has intensified competition between open-source and closed-source AI providers, particularly in enterprise coding applications.
Cost Is Becoming a Major Competitive Advantage
Databricks’ findings reinforce a broader industry trend: enterprises are increasingly evaluating AI models based not only on benchmark performance but also on the total cost of deployment.
| Enterprise Priority | Why It Matters |
|---|---|
| Model accuracy | Reliable code generation |
| Cost per task | Lower infrastructure spending |
| Context window | Better handling of large codebases |
| Open-source availability | Greater flexibility and customization |
| Deployment options | Reduced vendor lock-in |
As organizations integrate AI into software development at scale, inference costs have become a key factor in determining which models are deployed across engineering teams.
Chinese Open-Source Models Gain Global Adoption
Databricks is not alone in adopting Chinese open-source AI models.
According to recent industry reports, companies including Coinbase, Snowflake, and several AI startups have begun experimenting with or deploying Chinese models such as GLM 5.2, Kimi 2.7, and DeepSeek after finding that they offer competitive performance at significantly lower operating costs.
On AI model marketplace OpenRouter, Chinese open-source models have reportedly grown to account for more than 30% of weekly traffic, up sharply from the previous year.
A New Phase in the AI Race
The growing popularity of Chinese open-source models comes as geopolitical tensions increasingly shape the AI industry.
While the United States has tightened export controls on advanced AI chips and, at times, frontier AI models, China has rapidly expanded investment in domestic AI research. At the same time, Chinese authorities are reportedly considering restrictions on overseas access to some of the country’s most advanced open-source AI models, underscoring the strategic importance of AI technologies.
What It Means for the AI Industry
Databricks’ decision to adopt GLM 5.2 as its default coding engine signals a broader shift in enterprise AI adoption. Rather than choosing models solely based on headline benchmark scores, companies are increasingly prioritizing cost efficiency, deployment flexibility, and real-world performance.
For providers such as OpenAI and Anthropic, the move highlights growing pressure from rapidly improving open-source alternatives. If Chinese models continue narrowing the performance gap while maintaining significantly lower inference costs, enterprises may increasingly embrace open-source AI for everyday development tasks. The result could be a more competitive global AI ecosystem where pricing, efficiency, and openness become just as important as raw model capability.
Get the day’s top stories in your inbox
One concise email. No spam, unsubscribe anytime.
