AI Tools and Models: Which One Should You Actually Use?

AI is everywhere right now. Every week there’s a new model, a new tool, or a new “Copilot” for something. Blog posts, YouTube explainers, and endless Twitter threads are already tripping over each other to cover them all.

So why bother adding my own “me too” voice into the noise?

Well… not just for the sake of it. My goal here is to skip the buzzwords and glossy marketing and give you the sort of practical, real-world breakdown you’d actually use to decide: which AI tool should I fire up for this particular job?

Because here’s the secret: there’s no single “best” model. They’re like coworkers—some are great at spreadsheets, some are better at presentations, and a few just sit in the corner and argue with you.

You will notice that I did not include anything in here about using AI for images. That was by design. That could be a lengthy post all on its own, and not something that I intend to dive into.

OpenAI: ChatGPT

Models: GPT-4, GPT-4o, GPT-3.5
When to use it: general conversation, brainstorming, creative writing, explaining code to a human.
Strengths: Feels the most “chatty.” GPT-4o is multimodal, so you can throw images and text at it. Widely integrated (especially if you live in Microsoft Office all day). It can also pull in information from previous conversations
Weaknesses: Can still “hallucinate” answers like it’s being paid by the word. Not always fresh on current events.
Practical note: GPT-3.5 is great when you just need something quick and cheap; GPT-4o is where you go for heavier lifting. If you have access to it, GPT-5 is the most up to date

Anthropic: Claude

Models: Opus, Sonnet, Haiku
When to use it: analyzing giant documents, step-by-step reasoning, programming help.
Strengths: Handles absurdly long inputs without breaking a sweat. Tends to reason things out more carefully than ChatGPT.
Weaknesses: Sometimes too careful—hedges answers, adds disclaimers, or gets overly polite.
Practical note: Claude Sonnet is an especially good programming partner. It balances speed and depth really well, so it’s perfect for reviewing code, fixing bugs, or explaining why your function blew up.

Google DeepMind: Gemini

Models: Gemini 1.5 Pro, Gemini 1.5 Flash
When to use it: research, search-enhanced summaries, Google Workspace power-user life.
Strengths: Bakes directly into Gmail, Docs, and Sheets. Flash is lightning fast and cheaper than the heavyweights.
Weaknesses: Feels less flexible outside the Google bubble.
Practical note: If you’re already a Google person, Gemini will feel like an extra teammate. If not, it may just feel like that colleague who keeps forwarding you links.

GitHub Copilot

Copilot isn’t one thing—it’s two flavors of the same idea.

Copilot in your IDE (VS Code, JetBrains, etc.): autocomplete on steroids. It guesses the code you want as you type.
Copilot Chat: a side panel where you can ask questions, refactor code, or get explanations.

When to use it: anytime you’re actively writing code.
Strengths: Fast, context-aware, and doesn’t require switching out of your editor.
Weaknesses: Not really a generalist—it’s laser-focused on coding.
Practical note: I use Copilot for rapid scaffolding (think boilerplate, tests, repetitive functions), but if I want a real “second opinion” on design or debugging, I’ll ask GPT-4 or Claude Sonnet.

Cohere: Command R

When to use it: customer support, knowledge base Q&A, enterprise search.
Strengths: Retrieval-augmented generation (RAG) is its bread and butter.
Weaknesses: You’re not going to use it for poetry or D&D character sheets.
Practical note: Command R is like the reliable back-office worker—it may never trend on Hacker News, but it quietly keeps the company running.

Mistral / Mixtral

When to use it: when you want to self-host, fine-tune, or just avoid cloud lock-in.
Strengths: Lightweight, open weights, surprisingly efficient.
Weaknesses: Not as nuanced as GPT-4 or Claude.
Practical note: Great choice if privacy or control matters more to you than absolute top-tier accuracy.

Meta: Llama 3

When to use it: experiments, enterprise rollouts, custom assistants.
Strengths: Good mix of flexibility and performance. Open source, so you can tweak it to your heart’s content.
Weaknesses: Requires some setup and wrangling; not as “plug and play.”
Practical note: Think of Llama 3 as your AI toolkit—it’s not the prettiest out of the box, but if you want to build something tailored, it’s one of the best places to start.

Side-by-Side Examples

1. Programming Task

Prompt: “Write a function in JavaScript that checks if a string is a palindrome.”

Claude Sonnet:

function isPalindrome(str) {
  const cleaned = str.toLowerCase().replace(/[^a-z0-9]/g, '');
  return cleaned === cleaned.split('').reverse().join('');
}

GPT-4o:

function isPalindrome(str) {
  str = str.replace(/[^a-z0-9]/gi, '').toLowerCase();
  return str === [...str].reverse().join('');
}

GitHub Copilot:

function isPalindrome(str) {
  return str === str.split('').reverse().join('');
}

2. Creative Writing

Prompt: “Write the opening line of a noir detective novel.”

GPT-4o: The rain didn’t stop all night, and neither did the lies.
Claude Opus: By the time I lit my first cigarette, the city had already written its confession in smoke and shadow.
Gemini Pro: Downtown was still awake, but only the neon signs cared to admit it.

3. General Purpose / Everyday Help

Prompt: “What’s a quick, healthy lunch I can make in under 10 minutes?”

ChatGPT (GPT-4o): Try a Mediterranean wrap: hummus, spinach, cucumber, olives, and feta in a whole-wheat tortilla.
Claude Haiku: A bowl of quinoa (pre-cooked), cherry tomatoes, canned chickpeas, olive oil, and lemon.
Gemini Flash: Turkey and avocado on whole-grain toast + an apple.

So Which One Do You Pick?

Coding? Copilot for speed, then Claude Sonnet or GPT-4 for depth.
Summarizing documents? Claude Opus.
Brainstorming or writing? GPT-4o or Claude Opus.
Research? Gemini Pro.
Fast answers? Gemini Flash or Claude Haiku.
Privacy nerds? Mistral or Llama 3.

Wrapping It Up

The AI world is moving fast, but that doesn’t mean you need to chase every shiny new thing. Nor does it mean that AI will be replacing any of us anytime soon.

Instead, its like adding another thing into your toolbox that will enhance your workflow. You wouldn’t use a hammer to screw in a bolt, and you wouldn’t ask Copilot to write your next novel (unless you want something very… procedural).

In the next post, I’ll go beyond the “what tool to use” question and dig into the how: tips and tricks to consistently get better results, no matter which AI you choose.

Ramblings in time and measure