LLMs

LLM Categories?

Help people understand the capabilities of LLMs, to allow them to know what kinds of products they can build with them.

In understanding what AI can and can’t do, and how to turn your AI product into reality, it’s very useful to understand the different custom AI options that are currently available. The following is a brief summary of the current llm ecosystem, and how they can be used.

Understanding what AI is good at vs. not will help you save time and money!

• Prompting
• Fine-tuning
• RAG
• Agents
• Custom ml-models

How they Work

Example applications

Comparing LLM Functionalities

Method	Basic Functionality	Complexity	Data/Compute Req	Price	Customizability	Business Use Cases
Prompting LLM	Generate responses with carefully crafted prompts; fast, low effort.	★☆☆☆☆	★☆☆☆☆	$	★☆☆☆☆	General-purpose use: quick tasks, chatbots, content generation, brainstorming ideas.
Fine-Tuned LLMs	LLM is tailored to specific tasks or domains by training on additional data.	★★☆☆☆	★★★★☆	$$	★★★★☆	Domain-specific tasks: customer support, technical writing, product recommendations, research assistance.
LLM with RAG	Combines LLM outputs with retrieval from external data sources for real-time knowledge.	★★★☆☆	★★★☆☆	$$	★★★★☆	Knowledge-intensive tasks: legal queries, medical diagnostics, enterprise data retrieval, personalized recommendations.
LLM Agents	Use LLMs for multi-step reasoning and orchestration across tools/APIs.	★★★★☆	★★★★☆	$$$	★★★★☆	Complex automation: customer workflows, advanced research, multi-step decision-making, API orchestration.
Custom ML Models	Fully tailored model built from scratch for a specific task or domain.	★★★★★	★★★★★	$$$$$	★★★★★	Highly specialized tasks: fraud detection, unique domain applications, proprietary systems, custom business logic.

[P1] RAG vs. fine-tuning decision-tree