Zhipu AI's GLM 5.2 replaced Claude Opus 4.8 at Databricks at 34% lower cost

On July 11, 2026, Databricks announced a strategic decision: it is transitioning the open Chinese model GLM 5.2 from Zhipu AI to the status of primary tool for coding automation. In tests of its own multi-million line codebase, GLM 5.2 showed results identical to Claude Opus 4.8 from Anthropic, but 34% cheaper: $1.28 per task versus $1.94.

How Databricks Tested the Models

Databricks engineers ran code agents on the company's real production codebase — millions of lines of source code that reflect all the complexity and specificity of real development. This is critical because public benchmarks often measure models on synthetic tasks that don't match how the model behaves on your code specificity.

Databricks measured not just the quality and speed of generated code, but also practical cost at scale: how much money actually goes into coding automation per day, week, month in production. This is where the open GLM 5.2 showed its worth.

Results: GLM 5.2 Stands at the Same Level as Opus 4.8

The results are eloquent:

• GLM 5.2 from Zhipu AI — cost $1.28 per coding task
• Claude Opus 4.8 from Anthropic — $1.94 for the same task
• Savings when switching to GLM 5.2 — $0.66 per task, a 34% reduction in costs
• Quality of generated code — identical
• Speed of execution and accuracy — comparable
• Tests were conducted on Databricks' production data, not on public benchmarks

Databricks tested other candidates as well, but GLM 5.2 came out on top in the price-to-quality ratio.

Why This Disrupts Common Industry Belief

Traditional industry view: closed Western models (Claude Opus, GPT from OpenAI, Gemini from Google) automatically surpass open analogues from Asia. Databricks disrupts this myth with practice: on real production tasks, the open GLM 5.2 from Chinese Zhipu AI codes as well as proprietary Opus 4.8, but with notable savings.

"Companies often rely only on publicly available benchmarks and don't test models on their own data.

Results can differ significantly," — Databricks practically concludes.

The conclusion raises the main question: if open models compete with closed ones in practice, why pay 34% more?

What This Means for the Market

The AI model market is moving away from monopoly of one or two providers toward real diversity. Open models from Zhipu AI (GLM), Meta (Llama), Mistral are becoming competitive not just in price but also in quality on practical production tasks. For companies, this means: choice expands, there's no need to blindly take the model with the best public rating, you need to test on your own data and count the real cost of ownership.

Frequently Asked Questions

When Does GLM 5.2 Come Out in Public Access?

GLM 5.2 is already available as an open-source model from Zhipu AI. Any company can download, test, and deploy it on its own servers or in the cloud without licensing fees.

How

Much Will a Company Actually Save by Switching to GLM 5.2?

Savings depend on usage volume. If a company spends $100 per day on Claude Opus 4.8 for coding, switching to GLM 5.2 will give approximately $34 in savings daily or $12,400 per year. But this requires your own testing on your codebase.

*Meta is recognized as an extremist organization and is banned in the Russian Federation.