DeepSeek Prompting

Introduction

DeepSeek (R1, V3) is the rising open-weights model with strong reasoning and very low cost. Prompting style mirrors ChatGPT, with extra attention to reasoning prompts.

Beginner analogy: DeepSeek is the budget overachiever — cheap, capable, especially on math & code.

Understanding the topic

Core concepts to understand:

• OpenAI-compatible API — works with existing tooling.
• Strong on reasoning when triggered (CoT-friendly).
• Open weights — self-host for privacy.
• Pricing is <10% of OpenAI for many tasks.
• Pair with reflection prompts for accuracy boost.

Syntax reference

Visual workflow / architecture:

messages = [
  { role: "system", content: "You are a math tutor." },
  { role: "user",   content: "Solve step by step: 3x + 7 = 22" }
]

Real-world use

Many startups use DeepSeek for batch jobs and offline reasoning to slash AI cost.

Best practices

• Use CoT explicitly — DeepSeek rewards it.
• Pair with eval set; quality varies by task.
• Self-host for sensitive data.

Hands-on exercise

Interview preparation — practice these questions:

• Q1. Why use DeepSeek?
• Q2. Cost comparison?
• Q3. When NOT use DeepSeek?
• Q4. Open weights — implications?