Zero-shot prompting

Zero-shot prompting asks the model to perform a task with no examples — only the instruction and the input.

Zero-shot prompting relies on the model's pretrained capability rather than in-prompt demonstrations. It is the default style for large frontier models (GPT-4 class and above) on common tasks. Zero-shot is cheaper (fewer prompt tokens) and easier to maintain, but its quality is highly sensitive to prompt wording. "Translate the following to French" works fine; "Convert this to French in a casual Marseille accent" usually needs few-shot examples. As a rule of thumb: try zero-shot first, add CoT or few-shot only when evals show measurable gain.

When to use zero-shot prompting

• Standard tasks the frontier model has clearly seen in pretraining (translation, summarization).
• Latency- or cost-sensitive workloads.
• Prompts that are read by humans and need to stay short.

Common mistakes

• Assuming zero-shot will hit production quality without evals.
• Skipping zero-shot and jumping straight to fine-tuning — usually overkill.

FAQ

Related terms

• Few-shot prompting — Few-shot prompting supplies 2–10 input–output examples inside the prompt so the model imitates the pattern on a new input.
• Chain-of-thought prompting — Chain-of-thought (CoT) prompting tells a language model to write its reasoning steps before its final answer, increasing accuracy on multi-step problems.
• System prompt — A system prompt is the high-priority instruction block that defines a model's role, constraints, and default behaviors for an entire conversation.

Last updated: 2026-06-01. Raw markdown: https://promtable.com/glossary/zero-shot-prompting.md.