Custom vs. Pretarined Models
3 mins read - Created on Oct 23, 2025In Kimola, there are two main types of AI models you can work with: Pretrained Models and Custom Models. Both are designed to analyze text, but they differ in how they’re created, trained, and applied. Understanding the difference helps you choose the right approach for your research goals.
Sign in to your Kimola account and go to the Models section on the left panel. Select the Custom tab.
Custom Models
Custom Models are models that you create and train yourself — using your own data, categories, and labeling logic.
When you build a custom model, you teach the AI exactly how you want it to read and interpret text. This allows the model to learn the specific language, tone, and nuances of your dataset or industry.
If you collect customer feedback about coffee brands, you can train a model with categories like flavor, packaging design, price, or aroma. Once trained, your model will automatically recognize these themes in new feedback — even when customers use different words or expressions.
Because custom models are based on your own labeled data, they are highly accurate for domain-specific or business-specific research. They’re ideal when your project involves unique categories, specialized terminology, or internal data that general models wouldn’t fully understand.
Pretrained Models
Pretrained Models are ready-to-use AI models created and maintained by Kimola’s data team. They are trained on large, high-quality datasets and can understand general language patterns, emotions, and topics without any additional setup.
You can start using them immediately — no data labeling or training required.
The News Classifier automatically organizes articles into categories like politics, technology, or entertainment, while the Sentiment Classifier detects whether a comment is positive, neutral, or negative.
Pretrained models are great when you need quick, reliable insights from common types of text such as social media posts, reviews, or news articles. They’re optimized for general use cases and save time when you don’t need a fully customized setup.
You can explore all AI Models available in Kimola here
Which One Should I Use?
The best choice depends on your research goals and the type of data you’re analyzing.
Choose a Custom Model if:
- You have data from an industry not covered by Kimola’s pretrained, industry-specific models.
If your dataset belongs to a niche sector or a unique research field, a custom model can learn your terminology and context.
- You want the model to understand your own categories, tone, or business language.
- It learns from your labeled examples and delivers highly accurate, domain-specific results.
- You have unique datasets such as surveys, customer support tickets, or specialized feedback.
- A custom model captures the nuanced patterns that pretrained models might miss.
Choose a Pretrained Model if:
- You believe one of Kimola’s existing AI models already fits the type of data you want to analyze.
- Sentiment Classifier for detecting positive or negative opinions.
- Consumer Feedback Classifier for general product or business-related feedback.
- E-Commerce, Automotive, Banking, or Mobile Network Feedback Classifiers for industry-specific consumer conversations.
- Mobile App, Mobile Game, Software, Online Course, or HoReCa Feedback Classifiers for domain-focused user insights.
- News Classifier for organizing articles by topic.
- Hate Speech Detector for identifying discriminatory or harmful language.
- You want to start analyzing immediately without spending time on labeling or training.
- Pretrained models are ready to use right away — no setup required.
- You’re working with general text where common categories such as sentiment, topic, or entity extraction are sufficient.
- These models are designed for broad, cross-industry applications and deliver reliable, fast results.
