Back to Glossary
Prompting

Zero-shot Prompting

Definition

Zero-shot prompting is asking an LLM to perform a task without providing any examples, relying entirely on the model's pre-trained knowledge and the clarity of your instructions.

Why It Matters

Zero-shot prompting is the simplest and fastest approach - you just describe what you want without providing examples. Modern LLMs are remarkably capable at zero-shot tasks, making this the default starting point for most applications. It’s cheaper (fewer tokens) and easier to maintain than few-shot approaches.

How It Works

You provide only instructions without examples:

  • “Translate this text to French: [text]”
  • “Summarize this article in 3 bullet points: [article]”
  • “Classify this review as positive, negative, or neutral: [review]”

The model uses its training to understand the task and generate appropriate outputs.

When to Use

Start with zero-shot prompting for: straightforward tasks with clear instructions, simple classifications and extractions, tasks the model is likely trained on, and rapid prototyping. If zero-shot performance is insufficient, graduate to few-shot prompting by adding examples that demonstrate the desired behavior.