How to Use ChatGPT Astra: Access, Setup, and Getting Started

ChatGPT Astra is a model available in the ChatGPT model selector. To use it, you select it from the model picker at the top of the ChatGPT interface, the same way you switch between GPT-4o and other models. The access requirements vary by plan.

Before switching to a new model, a tool like fenbs keeps your project brief and accepted decisions in one place so you can continue working without rebuilding the context after the model change.

What "Astra" means in the ChatGPT context

The name Astra has also appeared in Google DeepMind's product work (Project Astra, their real-time multimodal assistant technology). If you encountered "Astra" first in the Google context, that is a different product from the model available in ChatGPT.

Within ChatGPT, Astra refers to a specific model you can select from the model picker. The details of what it does and which plans include it are maintained by OpenAI and change as the product evolves. Check OpenAI's model documentation↗ (opens in a new tab) and the ChatGPT help center for the most current information on what Astra offers and which plans include it.

How to find and select Astra in ChatGPT

1. Open ChatGPT and sign in.
2. Look for the model selector at the top of the chat interface. It usually shows the name of the currently active model (for example, GPT-4o).
3. Click or tap the model selector to open the list of available models.
4. Find Astra in the list. If it does not appear, your current plan may not include it.
5. Select Astra and start a new conversation. The model is now active for that conversation.

If Astra does not appear in your model list, check OpenAI's current plan comparison page to see which plans include it. Model availability can vary between free, Plus, Pro, and ChatGPT Work accounts.

What to check before switching to a new model

Switching to a new model in ChatGPT does not carry over the context from your previous conversations automatically, unless you have ChatGPT memory enabled and the relevant facts are in memory. For a task that depends on a project brief, you need to either provide the context at the start of the new conversation or have it stored in a project.

ChatGPT Projects store shared instructions that apply to every conversation in the project. If you use a Project for your work, the instructions carry across model changes within that project. Memory stores user-level facts that apply across all conversations.

Getting useful results from Astra

Start the session by providing the project context: what you are working on, the constraints that apply, and any decisions that are already settled. The more specific the context, the more relevant the first response. A model you have not used before may respond differently from the one you are used to, especially on tasks that are sensitive to tone or structure.

Run a short test on a known task before committing to Astra for important work. Use a task you have run on other models and compare the output against your criteria. This tells you whether the model's default behavior fits your workflow without risking an important project.

For a broader comparison of ChatGPT and Claude, see Claude vs. ChatGPT.

Astra versus other ChatGPT models

ChatGPT offers several models at once: GPT-4o, the reasoning-focused o-series models, and newer additions like Astra. Each has a different balance of speed, capability, and cost. Astra sits alongside these in the model selector; the right choice depends on what you are asking it to do.

For tasks that require current web information or multimodal input, check whether Astra has capabilities that differ from GPT-4o in those areas. For straightforward text tasks like writing, summarizing, or explaining, the difference between models is often small in practice. The clearest signal comes from running the same task on both and comparing the outputs against your criteria.

If you regularly switch models depending on the task, the main overhead is not the model change itself — it is re-establishing the context for each session. A session that starts from the full project context produces better results than one that starts cold. Building a habit of keeping that context in a consistent, retrievable form reduces the overhead of switching.

Keeping results consistent across sessions

ChatGPT memory and Projects both address the continuity problem. Memory stores user-level facts that carry across all conversations; Projects store shared instructions for a specific project. If you use a ChatGPT Project for a piece of work, those instructions apply regardless of which model you select within that project.

The limitation is that these features live inside ChatGPT. If you also use Claude, Gemini, or another assistant for parts of the same project, the context is split across different tools. For multi-tool workflows, keeping the source of truth in a neutral place gives you one record to update rather than several to synchronize.

How to do this with fenbs

1. Create a free account and add your project to fenbs.
2. Record the project context: the goal, the constraints, and any decisions already made. Keep it short enough to paste at the start of any new conversation.
3. Connect Claude and other supported apps from Integrations. For ChatGPT, connect in developer mode (check OpenAI's help pages for which plans include it).
4. When you switch to ChatGPT Astra for a task, paste or have the assistant read the current project context before starting. Ask it to save progress before the session ends.
5. Record any model-specific notes (what Astra handles well or differently) as context notes for future reference.