Genie code Customization

[Mailendiran]

Hi,

I use Genie code extensively for research , plan and development for building ETL scripts and code migrations.
As per my knowledge Databricks manages the backend LLM models for Genie code agent.

I wanted to try Genie code with Frontier models for my different use cases.
I could not find any article related to it.

If anyone have done it or have idea , please let me know.

Regards,
Mailendiran.

[irfanumar]

Databricks does not currently publish a supported list of selectable frontier models for Genie Code (e.g., choosing Claude Opus vs GPT-5 vs Gemini 2.5 Pro for a specific Genie Code session). I could not find any official article describing user-controlled model selection inside Genie Code.

However, there are AI functions in databricks, where you can select frontier models of your choice for different use cases. 

[Ashwin_DSA · Accepted solution]

Hi @Mailendiran,

From what’s publicly documented, Genie Code already uses frontier models behind the scenes, but it isn’t exposed as a bring-your-own-model or manual model-selection experience. Databricks describes Genie Code as an agentic system that routes tasks across multiple models and tools, automatically choosing the best model for the job, whether that is a frontier LLM, an open source model, or a custom model hosted on Databricks (Introducing Genie Code).

The public product documentation presents it the same way. The main Genie Code docs describe Genie Code as the AI coding and data assistant for Databricks workspaces and highlight capabilities like working across notebooks, SQL, Lakeflow, dashboards, and MLflow, along with personalisation through skills, instructions, and MCP servers, but they do not describe a customer-facing setting to choose a specific frontier model. The Use Genie Code page similarly explains Chat mode and Agent mode, and notes that if partner-powered AI features are disabled, Genie Code defaults to Databricks-hosted AI models.

So the short answer is that Genie Code does use frontier models, but as a managed Databricks experience rather than something you explicitly swap in and out yourself based on the current public docs. I also haven’t seen a public article that documents a supported way for customers to select a particular frontier model for different use cases such as ETL generation or code migration. If your goal is strict control over model choice, that would likely point more toward directly using foundation model or model serving capabilities on Databricks rather than Genie Code itself.

If this answer resolves your question, could you mark it as “Accept as Solution”? That helps other users quickly find the correct fix.