GenAI Fundamentals Cheat Sheet

Your complete reference for GenAI concepts, models, and parameters. Bookmark this page!

Model Quick Reference

Closed Source Models

Open Source Models

Parameter Tuning Guide

Temperature Settings by Task

Common Parameter Combinations

// Precise and factual (code, data extraction, math)
{
  temperature: 0,
  max_tokens: 2048,
  top_p: 1,
}

// Balanced (general conversation, explanations)
{
  temperature: 0.7,
  max_tokens: 1024,
  top_p: 1,
}

// Creative (brainstorming, writing, ideation)
{
  temperature: 1.2,
  max_tokens: 2048,
  presence_penalty: 0.6,
}

// Structured output (JSON, data extraction)
{
  temperature: 0,
  max_tokens: 1024,
  response_format: { type: "json_object" },
}

Token Quick Reference

1 token    ≈ 4 characters ≈ ¾ of a word
100 tokens ≈ 75 words
1K tokens  ≈ 750 words ≈ 1.5 pages
10K tokens ≈ 7,500 words ≈ 15 pages
100K tokens ≈ 75,000 words ≈ a novel

Estimating Costs

GPT-4o mini at $0.15/1M input tokens:
  1,000 short queries (~500 tokens each) = $0.075
  That's less than a penny for 1,000 queries.

GPT-4o at $2.50/1M input tokens:
  1,000 short queries (~500 tokens each) = $1.25

Claude 3.5 Sonnet at $3.00/1M input tokens:
  1,000 short queries (~500 tokens each) = $1.50

API Call Templates

OpenAI

const response = await fetch("https://api.openai.com/v1/chat/completions", {
  method: "POST",
  headers: {
    "Content-Type": "application/json",
    "Authorization": `Bearer ${process.env.OPENAI_API_KEY}`,
  },
  body: JSON.stringify({
    model: "gpt-4o-mini",
    messages: [
      { role: "system", content: "You are a helpful assistant." },
      { role: "user", content: "Your prompt here" },
    ],
    temperature: 0.7,
    max_tokens: 1024,
  }),
});
const data = await response.json();
const answer = data.choices[0].message.content;

Anthropic (Claude)

const response = await fetch("https://api.anthropic.com/v1/messages", {
  method: "POST",
  headers: {
    "Content-Type": "application/json",
    "x-api-key": process.env.ANTHROPIC_API_KEY!,
    "anthropic-version": "2023-06-01",
  },
  body: JSON.stringify({
    model: "claude-3-5-sonnet-20241022",
    max_tokens: 1024,
    system: "You are a helpful assistant.",
    messages: [{ role: "user", content: "Your prompt here" }],
  }),
});
const data = await response.json();
const answer = data.content[0].text;

Google (Gemini)

const response = await fetch(
  `https://generativelanguage.googleapis.com/v1/models/gemini-1.5-flash:generateContent?key=${process.env.GOOGLE_AI_KEY}`,
  {
    method: "POST",
    headers: { "Content-Type": "application/json" },
    body: JSON.stringify({
      contents: [{ parts: [{ text: "Your prompt here" }] }],
      generationConfig: { temperature: 0.7, maxOutputTokens: 1024 },
    }),
  }
);
const data = await response.json();
const answer = data.candidates[0].content.parts[0].text;

Key Terms Glossary

Decision Tree: Which Model Should I Use?

START: What am I building?
│
├─ Learning / experimenting?
│  └─→ GPT-4o mini (cheapest, good quality)
│
├─ Code-heavy application?
│  ├─ Need best quality → Claude 3.5 Sonnet
│  └─ Budget-conscious → GPT-4o mini
│
├─ Processing very long documents?
│  └─→ Gemini 1.5 Pro (1M+ tokens)
│
├─ High-volume production (100K+ requests/day)?
│  ├─ Quality critical → GPT-4o or Claude Sonnet
│  ├─ Good enough quality → GPT-4o mini or Claude Haiku
│  └─ Maximum control → Llama 3 via Groq/Together
│
├─ Data must stay private?
│  ├─ Can manage GPUs → Self-host Llama 3
│  └─ Need managed → Azure OpenAI or GCP Vertex AI
│
├─ Multimodal (text + images)?
│  ├─ Best quality → GPT-4o
│  └─ Budget → Gemini 1.5 Flash
│
└─ Simple classification / routing?
   └─→ GPT-4o mini or any small model

AI Prompt Templates for Learning More

Use these with any AI assistant to deepen your understanding:

Explain a concept: "Explain [concept] like I'm a developer who has never worked with AI. Use a coding analogy."

Compare technologies: "Compare [A] vs [B] for [use case]. Give me a table with pros, cons, pricing, and your recommendation."

Architecture advice: "I'm building [describe app]. Design the AI architecture — which models, what parameters, how to handle [specific challenge]."

Debug AI issues: "My AI feature is [describe problem]. Here's my prompt: [paste prompt]. Here's the output: [paste output]. What's wrong and how do I fix it?"

Optimize costs: "I'm spending $[amount]/month on AI API calls. Here's my usage: [describe]. How can I reduce costs without losing quality?"

Learn by building: "Walk me through building a [type of AI feature] from scratch. Start with the simplest version and then show me how to improve it step by step."

Common Patterns Reference

System Prompt Template

const systemPrompt = `You are a [role] specialized in [domain].

Your responsibilities:
- [Responsibility 1]
- [Responsibility 2]
- [Responsibility 3]

Rules:
- [Constraint 1]
- [Constraint 2]
- If you're unsure, say "I'm not certain about this."
- Always provide sources when citing facts.

Output format:
[Describe expected format — JSON, markdown, plain text, etc.]`;

Error Handling for AI Calls

async function callAI(messages: Message[], retries = 3): Promise<string> {
  for (let i = 0; i < retries; i++) {
    try {
      const response = await fetch("https://api.openai.com/v1/chat/completions", {
        method: "POST",
        headers: {
          "Content-Type": "application/json",
          "Authorization": `Bearer ${process.env.OPENAI_API_KEY}`,
        },
        body: JSON.stringify({
          model: "gpt-4o-mini",
          messages,
          temperature: 0.7,
        }),
      });

      if (response.status === 429) {
        // Rate limited — wait and retry
        const waitTime = Math.pow(2, i) * 1000;
        console.log(`Rate limited. Waiting ${waitTime}ms...`);
        await new Promise(resolve => setTimeout(resolve, waitTime));
        continue;
      }

      if (!response.ok) {
        throw new Error(`API error: ${response.status}`);
      }

      const data = await response.json();
      return data.choices[0].message.content;
    } catch (error) {
      if (i === retries - 1) throw error;
      console.log(`Attempt ${i + 1} failed. Retrying...`);
    }
  }
  throw new Error("All retries failed");
}

The Complete GenAI Developer Workflow

1. UNDERSTAND the problem — What AI capability do you need?
2. CHOOSE a model — Match capability to cost and requirements
3. DESIGN the prompt — System prompt + user prompt + examples
4. SET parameters — Temperature, max_tokens, response format
5. BUILD the integration — API calls, error handling, caching
6. TEST thoroughly — Edge cases, hallucinations, cost monitoring
7. OPTIMIZE — Adjust prompts, try different models, reduce tokens
8. MONITOR — Track costs, quality, latency in production

Keep Learning

This book covered the fundamentals. Here's where to go next:

• Prompt Engineering — Learn advanced techniques for getting better results
• Building with AI APIs — Hands-on projects with OpenAI, Anthropic, and Google
• RAG Systems — Build AI that uses your own data
• AI Agents — Build autonomous AI systems that use tools
• Fine-tuning — Customize models for your specific domain

Remember: The best way to learn GenAI is by building. Start with a simple project — a chatbot, a summarizer, a code helper — and iterate. Every project teaches you something new about how these models work and how to use them effectively.

Happy building!

---

🌐 www.genai-mentor.ai