AI Model Meaning: How a Model Differs From an AI App

The term "AI model" describes the computation engine that generates a response when you give it an input. The AI app — ChatGPT, Claude, Gemini, or Cursor, for example — is the product built around that engine. They are related but separate. Understanding the difference helps you know where your data lives, what memory actually means, and what changes when you switch tools. Tools like fenbs keep the work your team has agreed on outside any single model or app, so it travels with the project rather than staying locked in one product.

Most people use "AI" to mean the product they interact with: ChatGPT or Claude. Those products package a model with a set of tools, a user interface, memory features, and policies. The distinction matters when you want to understand what will change if you switch.

What a model is

A model is a large neural network — a mathematical function — trained on a large dataset. It takes input (your message, a document, an image) and produces output (a response, a summary, code). Training on data shapes the model's weights: numbers that determine how it responds to each kind of input. Those weights do not change after training.

The model has no awareness of you, no memory of your previous conversations, and no tools of its own. When a company says their product "uses GPT-4o" or "uses Claude 3.5 Sonnet," they mean that product sends your input to that model and returns its output.

What an AI app adds on top

The AI app is the software around the model. It includes the interface you interact with, the system prompt that shapes how the model behaves in that product, and features like web search, file upload, code execution, and image generation. The app also manages conversation history: it sends your earlier messages back to the model as part of each new request, which is why the model appears to remember what you said earlier in the same conversation.

Memory features — the kind that persist between conversations — also come from the app, not the model. ChatGPT's memory entries, Claude's Projects, and custom instructions are stored by the app and injected into new sessions. A different app using the same underlying model would not have access to them.

Context window: the model's working memory

Every model has a context window: the maximum amount of text it can process at once. Within that window, the model "knows" everything sent to it — the system prompt, previous messages, uploaded files, and your current question. Outside that window, it knows nothing about the conversation.

Context windows have grown significantly. Current leading models can handle hundreds of thousands of tokens, which is enough for long documents and extended conversations. But even a large context window empties at the end of the session. Anything not saved elsewhere is gone when the conversation closes.

Why the distinction matters for switching

When you switch from one AI app to another, you leave behind the memory features, chat history, files, and custom instructions that belong to the first app. The model's training weights are not the part that changes your experience day to day; the app layer is.

This is why keeping important work context in an outside tool matters. Decisions your team made, project rules, task status, and where things stand are usable by a new assistant only if they are somewhere that assistant can read. For how to choose between models, see how to choose an AI model.

Model versions and product names

AI companies release new model versions frequently and do not always keep older ones available. A product's name stays the same while the model behind it changes. ChatGPT today uses a different model version than ChatGPT two years ago. "ChatGPT" is the app; the model it uses is a specific version of GPT-4o or another OpenAI model.

When a capability difference matters — for a structured evaluation or a compliance review — note the exact model version, not just the product name. Product documentation usually lists the active model version in settings or release notes.

How to do this with fenbs

1. Create a free account on fenbs. Start with one project and write the goal, key rules, and current tasks into it.
2. Think of fenbs as the app layer that belongs to you: your decisions, your task state, your rules. The model is what your AI app provides; fenbs is what you provide.
3. Connect your AI apps from /integrations. When you switch from one app to another, the fenbs project context travels with you.
4. Record model decisions in Decisions and rules — "Using Claude for contract review as of October 2026" — so any connected assistant starts each session knowing the choice.
5. Before closing a session, ask the assistant to save Where we left off. The next assistant, on any model, reads that note before starting.

When fenbs is not needed

If your work stays inside one AI app and does not involve team members or recurring sessions, the app's own features — Projects, custom instructions, memory — may be enough. fenbs is useful when the work needs to survive a model change, an app switch, or a handoff to another person or AI session. fenbs does not host or train models and does not import chat histories from other apps.