SnowPro Specialty: Gen AI Question 230
Single answerCortex LLM Playground (Public Preview)A data science team is evaluating several large language models for an internal customer-support summarization use case. They want a fast way to compare how different prompts and models behave on the same sample ticket text before moving to a production implementation in SQL or Python. They also want to avoid building a custom app for this early testing phase. Which Snowflake feature is the best fit for this requirement?
- A
Use Cortex LLM Playground in Snowsight to interactively test prompts across supported models on sample inputs.
- B
Create a dynamic table that automatically compares model responses and ranks prompt quality over time.
- C
Use Snowpipe Streaming to send prompts to multiple models and visually compare their outputs in Snowsight.
- D
Build a Streamlit in Snowflake app first, because Cortex LLM Playground is only intended for production deployment workflows.
Show answer and explanation
Correct answer: A
Explanation
The best answer is Cortex LLM Playground in Snowsight. In public preview, it provides a no-code or low-friction environment to experiment with prompts and compare supported LLM outputs interactively, which is ideal for early evaluation and prototyping. This aligns with Snowflake best practices of validating prompt behavior and model choice before embedding LLM calls into production SQL, Python, or application workflows. Features such as dynamic tables and Snowpipe Streaming serve different purposes: transformation orchestration and ingestion, respectively. Streamlit in Snowflake is valuable when a team needs a custom user experience, but for quick model and prompt exploration, Cortex LLM Playground is the more appropriate starting point.
- A. Correct.
Correct. Cortex LLM Playground is designed for interactive experimentation in Snowsight, especially during early-stage prompt and model evaluation. It lets users quickly try prompts against supported LLMs without first building a custom interface or production workflow. This makes it well suited for comparing model behavior on the same business text, such as support tickets, before operationalizing the solution.
- B. Incorrect.
Incorrect. Dynamic tables are used for incremental data transformations and pipeline-style processing, not for interactive LLM prompt experimentation or side-by-side model testing in a UI. Someone might choose this if they confuse data pipeline automation with model evaluation tooling.
- C. Incorrect.
Incorrect. Snowpipe Streaming is used for low-latency data ingestion into Snowflake, not for prompt execution, model comparison, or LLM experimentation. This distractor reflects a common misconception that all real-time Snowflake features are interchangeable for AI workflows.
- D. Incorrect.
Incorrect. Streamlit in Snowflake can be useful for building custom applications, but it is not required for initial prompt testing. Cortex LLM Playground exists specifically to support rapid interactive exploration without needing to build an app first. The statement that the Playground is only for production deployment workflows is inaccurate.