🎯Choosing the Right Model

When to use Opus vs Sonnet vs Haiku — cost, speed, and quality trade-offs

Choosing the Right Model

One of the most important decisions when building AI applications is choosing the right model. Each Claude model has different strengths, speeds, and costs.

Available Claude Models

Claude Opus 4 — The Most Powerful

Best for: Highly complex tasks, deep analysis, advanced coding, research
Speed: Slowest
Cost: Highest ($15/$75 per million tokens)
Context window: 200K tokens
When to use:
- Analyzing complex, multi-document tasks
- Writing complex architectural code
- Tasks requiring multi-step reasoning
- Academic research and analysis

Claude Sonnet 4 — The Perfect Balance

Best for: Most everyday tasks, coding, writing, analysis
Speed: Fast
Cost: Medium ($3/$15 per million tokens)
Context window: 200K tokens
When to use:
- Building chatbots and interactive apps
- Writing and reviewing code
- Summarizing documents
- Most production use cases

Claude Haiku 3.5 — Fastest and Cheapest

Best for: Simple, fast tasks, high volume
Speed: Fastest
Cost: Lowest ($0.80/$4.00 per million tokens)
Context window: 200K tokens
When to use:
- Text classification
- Simple Q&A
- Structured data extraction
- High-volume, budget-conscious applications

Comprehensive Comparison Table

Criteria	Opus 4	Sonnet 4	Haiku 3.5
Quality	⭐⭐⭐⭐⭐	⭐⭐⭐⭐	⭐⭐⭐
Speed	⭐⭐	⭐⭐⭐⭐	⭐⭐⭐⭐⭐
Cost	$$$$$	$$$	$
Coding	Excellent	Excellent	Good
Reasoning	Best	Very Good	Basic
Writing	Best	Excellent	Good
Latency	10-30s	3-10s	1-3s

Decision Tree

Use this flowchart to choose your model:

Is the task complex and requires deep reasoning?
├── Yes → Is budget flexible?
│   ├── Yes → Use Opus 4
│   └── No → Use Sonnet 4 with extended thinking
└── No → Do you need high speed or high volume?
    ├── Yes → Use Haiku 3.5
    └── No → Use Sonnet 4

Practical Cost Comparison

Same task (summarizing a 2,000-word article) on each model:

Model	Input Tokens	Output Tokens	Cost	Latency
Opus 4	~3,000	~500	$0.0825	~15s
Sonnet 4	~3,000	~500	$0.0165	~5s
Haiku 3.5	~3,000	~500	$0.0044	~2s

Opus is 19x more expensive than Haiku for the same task!

Use Case Examples

Use Opus when:

Building an automated code review system that needs deep architectural understanding
Analyzing complex legal contracts
High-quality research writing

Use Sonnet when:

Building a customer support chatbot
Translating documents
Building productivity tools

Use Haiku when:

Classifying thousands of messages (positive/negative/neutral)
Extracting data from invoices
Building a simple auto-reply system

Switching Models in Code

JavaScript
// Super easy — just change the model name
const model = "claude-sonnet-4-20250514"; // or claude-opus-4-20250514 or claude-haiku-3-5-20241022

const response = await client.messages.create({
  model: model,
  max_tokens: 1024,
  messages: [{ role: "user", content: "Your question here" }],
});

Python Example

Python
import anthropic

client = anthropic.Anthropic()

# Easy to switch between models
model = "claude-sonnet-4-20250514"

response = client.messages.create(
    model=model,
    max_tokens=1024,
    messages=[{"role": "user", "content": "Your question here"}],
)

print(response.content[0].text)

Pro Tip: Model Routing

Many successful applications use different models for different tasks:

JavaScript
async function smartRoute(task, complexity) {
  let model;

  if (complexity === "high") {
    model = "claude-opus-4-20250514";
  } else if (complexity === "medium") {
    model = "claude-sonnet-4-20250514";
  } else {
    model = "claude-haiku-3-5-20241022";
  }

  return await client.messages.create({
    model,
    max_tokens: 2048,
    messages: [{ role: "user", content: task }],
  });
}

Advanced: A/B Testing Models

JavaScript
async function abTestModels(prompt) {
  const models = [
    "claude-opus-4-20250514",
    "claude-sonnet-4-20250514",
    "claude-haiku-3-5-20241022",
  ];

  const results = await Promise.all(
    models.map(async (model) => {
      const start = Date.now();
      const response = await client.messages.create({
        model,
        max_tokens: 1024,
        messages: [{ role: "user", content: prompt }],
      });
      const latency = Date.now() - start;

      return {
        model,
        latency,
        inputTokens: response.usage.input_tokens,
        outputTokens: response.usage.output_tokens,
        response: response.content[0].text,
      };
    })
  );

  results.forEach((r) => {
    console.log(`Model: ${r.model}`);
    console.log(`Latency: ${r.latency}ms`);
    console.log(`Tokens: ${r.inputTokens} in, ${r.outputTokens} out`);
    console.log("---");
  });

  return results;
}

Quick Summary

Opus 4: Best quality, slowest, most expensive — for critical tasks only
Sonnet 4: Perfect balance — the default choice for most applications
Haiku 3.5: Fastest and cheapest — for simple tasks and high volume
Start with Sonnet, upgrade to Opus only when needed, downgrade to Haiku when possible

Next: We'll learn how to use Claude Code in your Git workflow.

Module 2

4/6

📡 Your First API Call

Anthropic Console Walkthrough 🖥️

4/6