Choosing the Right AI Model Provider for Your Business - Part 1 OpenAI

Picking the right AI model used to be easy—ChatGPT was the go-to. But now, it feels like every month a new model drops, and keeping up is getting harder. We’ll try to break down what’s available, the pros and cons of different options, and help you figure out which model makes sense for what you’re working on. Whether you’re building tools for your business, generating content, or writing code, this will help you choose the right model without the headache.

Two Types of AI Models: Open Source vs. Closed Source

Before we dive into the specific models, there are two broad categories you need to know about:

Open Source Models: Weights available. You can download it, tweak it, and run it however you want. Think of Meta’s Llama 3 or Mistral. You get full control, but it also means you’re responsible for running and maintaining them yourself.
Closed Source Models: These are behind a company’s walls. You get access through APIs, but you don’t see the internals. Big players here include OpenAI’s ChatGPT and Anthropic’s Claude and Google’s Gemini You don’t need to worry about the technical setup—just plug in and start using them.

Deep Dive: OpenAI (GPT-4o + o1 )

Let’s start with OpenAI since it’s probably the name you’ve heard the most. They’ve got several models in play, but we’ll focus on the most useful ones and how to choose between them.

GPT-4 Series

This is OpenAI’s main offering. But even within GPT-4, there are a few versions that are constantly getting updated.

GPT-4o: High-intelligence flagship model for complex, multi-step tasks
GPT-4o (2024, 05, 08): OpenAI updates these models frequently, usually improving performance or adding features. This is the most recent checkpoint of gpt-4o. While not specifically stated, we reccoemdn using this one because its cheaper than GPT-4o, while also being better.
ChatGPT-4o Latest: This is the version used in ChatGPT (what you’re using right now). It’s particularly good at creative tasks like writing blog posts or generating marketing copy.

GPT-4 Mini

GPT-4 Mini is the cheaper, faster sibling. It’s not as smart as the full GPT-4 models, but it’s quick and perfect for tasks that don’t require deep reasoning, like classifying text or summarizing content. It’s the model you’ll want when you need high volume and low cost.

Why you’d use it:

If you need speed and don’t want to break the bank.
For simple outputs like summaries or light automation.
Doing straight forward grunt work

o1 Preview and o1 Mini

These models are newer and more focused on reasoning. The difference is that they need more detailed context to work well. Think of the o1 series as giving instructions to a senior developer: if you give them the full background, they’ll make smart decisions. If you don’t, it’ll just feel slow and expensive.

OpenAI Strengths and Weaknesses

Where OpenAI shines:

Instruction following: You ask it to do something, and it’ll follow your directions well.
Structured outputs: Great for generating JSON or structured data formats, which developers love.

Where OpenAI struggles:

Creativity: Some versions (like GPT-4 Mini) are less creative than competitors.
Latency: With heavy workloads, like summarization at scale, you might experience delays.

How to Pick an OpenAI Model

For fast and cheap tasks: GPT-4 Mini is a good option. It handles a lot without costing you much.
For creative or general tasks: ChatGPT-4 Latest is your go-to, especially for things like content generation or conversational AI.
For complex or technical tasks: If you need precision, especially for structured outputs, go with one of the full GPT-4 versions (2024/05/08, for example).
For in-depth analysis, high level tasks, and difficult and challenging problems: Use the o1 models. They think and take more problem to tackle a problem. Although we use them in chatGPT and not the API - its’ more cost effective!

OpenAI Models Overview and Pricing

Model	Context Size	Version	Input Token Price (per 1M)	Output Token Price (per 1M)	Batch API Input Token Price (per 1M)	Batch API Output Token Price (per 1M)
GPT-4o	128K	2024-08-06	$2.50	$10.00	$1.25	$5.00
GPT-4o	128K	2024-05-13	$5.00	$15.00	$2.50	$7.50
GPT-4o mini	128K	2024-07-18	$0.150	$0.600	$0.075	$0.300
o1-preview	128K	2024-09-12	$15.00	$60.00*	N/A	N/A
o1-mini	128K	2024-09-12	$3.00	$12.00*	N/A	N/A

*Batch API pricing offers a 50% discount if requests are submitted as a batch. Responses are returned within 24 hours. *Output tokens for o1-preview and o1-mini include internal reasoning tokens generated by the model that are not visible in API responses.

Thinking tokens can get very very expensive. You are not billed these in chatGPT which is why we recommend using it within the app

Wrapping Up OpenAI

In short, OpenAI’s models are reliable and versatile. They handle a wide range of tasks well and are especially good when you need structured, predictable results. If your project involves handling data or building apps, they’re a safe bet.

Next Up: Anthropic’s Claude and Google’s Gemini

In the next blog, we’ll cover Anthropic’s Claude 3.5 Sonnet, which is great for coding, and Google’s Gemini 1.5 Flash and Pro, which has quietly gotten a lot better lately. We’ll compare how they stack up to OpenAI and where each one really shines.