Custom AI Models vs. Pre-Trained ChatGPT: What’s Best for Your Business?

Fine-tune AI Model

1. Go to and create an account.

2. Once logged into your account, go to Dashboard

3. Now click on Fine-tuning from the left side menu.

4. From the right side, click Create

5. Here you need to provide some information:

• Method: Supervised
• Base Model: Use the latest one (Recommended)

In this context, 1 token approximately equals 1 word, but some words may be split into multiple tokens. For example, in the sentence: "FireVPS is the best Windows RDP provider,"

There are 7 words, and it would typically count as 7 tokens, as each word is treated separately. However, in some cases, common phrases or complex words may be split differently based on the tokenizer.

If your training dataset contains 1 million tokens, the pricing mentioned above will be affected accordingly.

6. Next step is to upload the training data set. For that we need to create a new training data set in Jsonl (Json in single line). Let’s assume we wanted to train ChatGPT on FireVPS business and we want to create content. Here are step by step for process for creating data sets:

A. Collect all previous FireVPS post from social media like LinkedIn, Facebook, Twitter etc and Save it in the notepad. And we will mark this as #post. This should look like this:

B. Now, we need to create a PROMPTfor these posts. We can either write it ourselves or use ChatGPT for assistance. To do this effectively, we must reverse-engineer the process.

C. Your Notepad should look like this

D. Like this we have to create 50-80 sets of #PROMPT and #POST. Once our data collection is ready now it is time to format this in Jsonl.

E. Open Visual Studio Code and Paste below format

{"messages":[{"role":"user","content":"#PROMPT"},{"role":"assistant","content":"#POST"}]}

F. Now replace the #PROMPT and #POST with the content. Your file should be look like this

G. Now convert all 50-80 sets of data to this Json format and save the file as JSON Lines

H. Now your training Data Sets are ready to upload.

7. Now upload the training data set and leave all other fields as it is.

8. Now click on Create.

Congratulations! You have successfully fine-tuned your AI model. From now on, whenever you ask or request content, it will generate responses based on your dataset, matching your way of speaking and tone.

Create Brand Brief

We need another important piece of information: The Brand Brief.

So, what is a Brand Brief? It’s a document that ensures our fine-tuned AI model generates content aligned with our brand's policies, refund terms, target audience, objectives, and more.

By providing a clear Brand Brief, we can make sure the AI produces content that truly represents our brand. Below is a Brand Brief format to guide you.

• Brand Name
• Company Background
• Product/Service Vision Statement
• Product/Service Mission Statement
• Key Themes & Topics
• Objectives
• Target Audience
• Unique Value Proposition (UVP)
• Dislike and Opponents
• Associate product and services
• Tone of Voice

This is a general format for a simple brand brief. Providing more accurate and detailed information will improve the content quality.

If you're finding it difficult to create one yourself, you can use ChatGPT for assistance. Simply provide your company information and ask it to generate a brand brief for you.

Now Test the System

1. Go to Playground

2. Select your Fine-tune AI model in Model Section

3. In System message paste your Brand Brief

4. Click on Additional Message

5. In the User section give your prompt. Example: Based on the given Brand Brief, write content about "Why dedicated server is better than shared hosting"

6. Now press Enter. Content will be generated.