What AI models does ChatGPT use?

ChatGPT uses OpenAI's models, including GPT-4o for general-purpose tasks and the o-series reasoning models for complex logic and math. Which models are available depends on your plan. Check your ChatGPT account settings or OpenAI's documentation for current options.

What is the difference between GPT-4o and the o-series?

GPT-4o is faster and suitable for most interactive work: writing, research, analysis, and coding. The o-series models take longer to respond because they reason through problems step by step. They are stronger on complex logic and multi-step problems but are overkill for simpler tasks.

Does ChatGPT remember things between conversations?

ChatGPT can store memory entries about you that persist across conversations, but this is a feature of the app, not the model. A different AI assistant does not have access to that memory. For shared project context, an outside tool like fenbs is needed.

Can I switch ChatGPT models mid-project?

Yes. ChatGPT lets you switch models within the app. The chat history and custom instructions in ChatGPT carry over within the app. If you switch to a different AI product, you need to provide the project context another way.

What is the ChatGPT Work or team plan?

OpenAI has updated its product names over time. For current plan names, features, and which models they include, check OpenAI's official pricing and plans page directly, as names and features change.

ChatGPT AI Models: Choose One for the Work You Need

ChatGPT gives you access to several models and modes. This guide explains how they differ and how to choose based on your actual tasks rather than the highest tier available.

August 19, 2026Updated October 21, 20266 min read

ChatGPT is an AI assistant from OpenAI that lets users access different underlying models depending on their plan and task. The model running a conversation matters: it determines what the assistant can do, how it reasons, and how fast it responds. Tools like fenbs keep your project brief and progress outside the chat, so you can try a different model or reconnect after a break without rebuilding the context from scratch.

Understanding what the different ChatGPT models and modes actually do helps you choose the right one for a given task rather than always defaulting to the highest tier.

The main ChatGPT model families

ChatGPT runs on OpenAI's models, which fall into a few families. GPT-4o is the main general-purpose model: fast, multimodal, and capable across writing, research, analysis, and coding. It handles image inputs and file uploads on most plans. It is the default for most interactive tasks.

The o-series models — such as o1 and o3 — are reasoning models. They take longer to respond because they work through a problem step by step before producing output. They perform better than GPT-4o on complex logic, math, and structured analysis, but they are overkill for a simple writing or research task.

Smaller or earlier model versions, including GPT-4o mini, are faster and cheaper per token. They are suitable for high-volume tasks where the quality requirements are modest: simple classification, short answers, or routing. For detailed current model availability by plan, see OpenAI's documentation (opens in a new tab).

Modes and tools that change how ChatGPT works

The model is one part of how ChatGPT behaves. The mode and tools available in a session are another. Web browsing lets ChatGPT search the current web, which is useful for recent information but adds a step that can introduce errors. Code execution lets it run Python to solve math problems, process data, or generate charts. Image generation connects to DALL-E.

Custom instructions and Projects in ChatGPT let you store persistent context inside the app. Custom instructions tell the model something about you that applies to all conversations. Projects collect files, instructions, and conversation history for a specific piece of work. Both are useful — but both stay inside ChatGPT. If you switch models or move to another assistant, that context stays behind.

How to choose a model for your task

For most interactive work — writing, editing, research questions, short analysis — GPT-4o is the practical default. It is fast enough for conversation and capable enough for the large majority of knowledge-work tasks.

Switch to a reasoning model when the task involves multiple logical steps that must all be correct: a complex code problem, a structured argument that must hold together, or a math proof. The slowdown and higher cost are usually worth it only when accuracy on the whole chain of steps matters.

Use a smaller model when you are running a high volume of simple tasks through the API or a custom tool, and cost per task is a constraint. For direct work in the ChatGPT interface, the plan you are on determines which models you can choose; check your plan settings.

How fenbs helps

Keep context reusable when you change models

When you move from one ChatGPT model to another, or switch to a different AI app, fenbs holds the project brief, current task state, and Decisions and rules. The next session starts with the same context, whatever model is powering it.

Create a free account

Free for one person · No credit card · Works with Claude, ChatGPT, Cursor and more

What ChatGPT keeps and what it does not

ChatGPT retains chat history, memory entries, custom instructions, and project files inside the app. This context is not automatically available in other AI tools, and it is not easily portable. If you stop using a project in ChatGPT, the history is there when you return, but another assistant starting fresh has none of it.

Memory in ChatGPT is separate from the model. ChatGPT can remember facts about you across conversations, but that memory belongs to the app layer, not the model. A different model in ChatGPT uses the same memory; a different assistant outside ChatGPT does not.

How to do this with fenbs

Create a free account on fenbs. Add your current project with a description of the goal and the key rules for the work.
Connect ChatGPT to fenbs from /integrations. Developer mode in ChatGPT is required; check OpenAI's help page for which plans include it.
At the start of each ChatGPT session, ask it to read the fenbs project context: the rules, current tasks, and Where we left off.
When you finish a session, ask it to save progress back to fenbs before closing the chat.
If you switch to Claude or another supported assistant, connect it the same way. It reads the same project record and continues from where ChatGPT left off.

What changes between ChatGPT model versions

OpenAI updates its models regularly. Behavior can shift between updates — a style or tone you relied on may change slightly with a new version. If your workflow depends on consistent output format, test each new model version before relying on it in production. Note which version was active when you made a model decision so a future re-check starts from a known point.

ChatGPT's settings usually show the active model version in the conversation. Checking it takes ten seconds and makes your decision records far more useful to the next person who reads them.

When fenbs is not needed

If your work is self-contained in a single ChatGPT conversation and does not involve team members or recurring decisions, ChatGPT's built-in Projects feature may be enough. fenbs adds value when the context needs to survive model changes, app switches, or handoffs between people and AI sessions. fenbs does not import ChatGPT chat histories.