Why GPTs Are No Longer the Future and you must use Projects : A Strategic Wake-Up Call for Individuals and Institutions.

While OpenAI's GPT ecosystem has revolutionized how educators, students, and instructional designers engage with AI, a silent crisis is emerging. Most “Custom GPTs”, those handy bots built using GPT Builder, are still running on the now-outdated GPT-4-turbo model from 2023. And nobody talks about it!

This is a strategic turning point that few are discussing but that you cannot afford to ignore. As more advanced reasoning models like GPT-4o, GPT-4.1, and the O-series reshape the landscape of intelligent systems, clinging to GPT-4-turbo while using your "GPTs" is no longer just a minor oversight, it’s an operational liability. OpenAI GPTs on ChatGPT don’t allow you to change the target model used to process your requests. The problem is that you can’t see which engine is behind it... and the LLM behind GPTs is still GPT-4-turbo, which is quite outdated.

Instead, OpenAI now offers the "Projects" feature for paid users. This is much more powerful because for each project's request, you can choose the specific model (such as GPT-4, GPT-3 series, GPT-4.1, and others) that will handle your question.

Nothing now prompts us to continue using "GPTs" with the outdated GPT4-Turbo model while "Projects" use the latest OpenAI models. Yes, and No. Because beyond the aspect of the language models used, GPTs also offer other features like sharing and collaboration, which many "Projects" offer much less. We will see this further on.

A Brief History of GPT Model Evolution (2023–2025)

To understand the magnitude of this transition, let’s briefly revisit the evolution of OpenAI’s flagship models:

• GPT-4 (March 2023): The original GPT-4 introduced massive improvements in reasoning, multilingual capabilities, and fewer hallucinations than GPT-3.5.
  

• GPT-4-turbo (Nov 2023): A cheaper, faster variant of GPT-4, used in most ChatGPT Pro accounts and Custom GPTs. However, OpenAI never clarified if it's functionally the same as GPT-4.
  

• GPT-4o (May 2024): "o" for "omni"—this was a leap forward. GPT-4o supports seamless multimodal reasoning (text, image, audio, and soon video) and exhibits fluid conversation, contextual recall, and live perception.
  

• GPT-4.1 and the O-Series (o3, o3-mini, o4, o4-mini, o4-mini-high) (2025): Introduced behind the scenes in ChatGPT’s “More Models” feature, these models have expanded context windows, improved planning and memory, and far superior cognitive architecture.
  

Key Takeaway: All GPTs are still using GPT4-Turbo, and you cannot change that. GPT-4-turbo was powerful in 2023. In 2025, it's the equivalent of using a smartphone from 2016 in a world of AR interfaces and intelligent agents.

"Old" GPT-4-Turbo (with GPTs) vs. GPT-4o (With Projects): A Cognitive and Technical Comparison

Feature	GPT-4-turbo (2023)	GPT-4o (2024)
Reasoning Ability	Good, but linear and error-prone in abstraction	High-level symbolic, analogical, and multi-hop reasoning
Context Window	128k tokens	128k tokens
Multimodality	Limited (mainly text + image)	True multimodality (text, image, video, audio)
Memory Support	None in Custom GPTs	Yes, in ChatGPT Projects
Adaptability	Fixed personality and instructions	Dynamic, modular, and stateful behavior
Cost Efficiency	Lower compute cost	Slightly higher but more efficient and accurate
Live Interaction	Delayed and transactional	Near-real-time, proactive, and perceptive

Pedagogical Impact Example:

• A GPT-4-turbo Custom GPT (GPTs) can answer “What is the Kolb Learning Cycle?” accurately.

• GPT-4o, integrated within ChatGPT Projects, can adaptively coach a student through each phase based on real-time feedback, mood (via audio), or learning behavior.

We always recommend using the perfect Prompt structure by following at least our Prompting 101 Lecture series. Visit the "Prompting 101 Lectures Series" Introduction Page.

For even better results, we invite you to craft your prompts using the UP Method (University 365 Prompting). This approach encourages you to create separate, reusable personal CONTEXT, PERSONA, and ROLE files for each situation. You can "convoque" these files when appropriate by uploading them in the prompt or within the project. Discover the UP Method.

Why GPTs Are No Longer the Future for Education and Beyond

Despite their initial utility, here are the reasons academic stakeholders should reconsider using GPTs built with GPT Builder:

1. Cognitive Inflexibility

GPT-4-turbo struggles with:

• Non-linear abstraction

• Exploratory thinking

• Adaptive dialogue continuity

This weakens its utility in complex learning tasks like Socratic questioning, cognitive scaffolding, or design thinking workshops.

2. No Stateful Memory

GPTs do not retain context across sessions, making them incapable of true mentorship or longitudinal feedback loops. In contrast, ChatGPT Projects and GPT-4o support memory and session continuity.

3. Limited Personalization

GPTs have rigid personality structures based on static instructions. Modern pedagogical AI demands dynamic user modeling that GPT-4-turbo cannot deliver.

4. Technical Obsolescence

Educational AI must evolve with tech standards. GPTs stuck on GPT-4-turbo lack integration with updated APIs, modules, and multi-agent ecosystems emerging with GPT-4.1 and GPT-4o.

5. Strategic Lock-In Risk

Investing in GPTs today is building legacy tools with shrinking relevance. Future integration and migration to Projects will require additional effort and cost.

The Rise of ChatGPT Projects

The Next Chapter in Applied AI

The Projects feature in ChatGPT (launched in 2025) is more than a UI update—it’s a paradigm shift:

What Makes Projects Superior?

• Stateful AI Agents: These agents remember users’ goals, past interactions, and context.
  

• Access to Advanced Models: Projects allow use of GPT-4o, GPT-4.1, and future O-series agents.
  

• Modular Architecture: Projects can include tools, files, databases, and UI components for rich AI workflows.
  

• Expanded Token Memory: Like Claude 3 Opus (200K tokens) or Gemini 1.5 (1M+), GPT-4o within Projects supports deep, structured thinking—ideal for research or curriculum design.
  

GPTs vs Projects Beyond The Mear LLM Choice,

A Complete Different Approach

OpenAI's current offerings present a clear distinction between Custom GPTs and ChatGPT Projects, each with unique capabilities and limitations.

Custom GPTs (using sadly onvly GPT-4-turbo) are designed for ease of sharing with a large audience and collaboration. They can be published in the GPT Store, allowing users to discover and utilize them without needing insight into their internal configurations. This shareability makes them accessible tools for a broad audience. But for many users, they refer to it as "black boxes" because they lack full control over the instructions and the training data.

ChatGPT Projects, on the other hand, are private workspaces tailored for individual users. They offer advanced features such as stateful memory, modular architecture, and access to multiple models, including GPT-4o, the o-series (o3, o4) and even the new GPT-4.1. However, Projects are not shareable in the same manner as Custom GPTs; they are confined to the creator's workspace and cannot be published or accessed by others in the GPT Store.

As of now (May 2025), OpenAI has not announced plans to integrate the shareability of Custom GPTs with the advanced functionalities of ChatGPT Projects. The company continues to develop tools aimed at simplifying the creation of agentic applications, such as the new Responses API and Agents SDK, which may influence future capabilities.

For users seeking both shareability and advanced features, third-party platforms like Eden AI's AskYoda offer alternatives. AskYoda allows users to select from various large language models and integrate diverse data sources, providing a customizable and shareable AI experience. Eden AI

In summary, while Custom GPTs and ChatGPT Projects each serve distinct purposes within OpenAI's ecosystem, there is currently no unified solution that combines the shareability of GPTs with the advanced capabilities and LLM choice of Projects. Users must still choose the tool that best aligns with their specific needs and objectives, knowing perfectly the limits of each one.

Shareability: GPTs vs. Projects

Custom GPTs:

• Public Sharing: Can be published in the GPT Store, making them discoverable and usable by any ChatGPT user.
  

• Private Sharing: Can be shared via direct links, allowing specific users access without public listing.
  

• Team Collaboration: In ChatGPT Team workspaces, GPTs can be shared among team members, facilitating collaborative use.

ChatGPT Projects:

• Individual Use: Designed primarily for personal organization of chats, files, and custom instructions.
  

• Limited Sharing: Currently, Projects are not shareable with other users, even within the same team workspace.
  

• Feature Requests: Users have expressed interest in shared project memory and collaborative features, but these are not yet implemented.

Example Use Case in Education

Adaptive Curriculum Assistant

With GPTs: You get templated answers to “Create a lesson plan on AI ethics.”

With Projects: The assistant builds adaptive lesson sequences, tracks learner behavior, adjusts content difficulty, and updates materials based on the latest research pulled from integrated data sources.

Comparative Lens

Platform	Projects (OpenAI)	Claude 3 (Anthropic)	Gemini 1.5 (Google) Gems
Model Access	GPT-4o, GPT-4.1, O-series	Claude 3 Opus	Gemini 1.5 Pro
Memory	Persistent, editable	Contextual but ephemeral	Long-context, temporary memory
Tool Integration	Advanced (browser, code, files)	Limited	Strong multimodal, workspace-based
Usability	Intuitive & modular	Research-focused	Enterprise-ready, complex
Academic Alignment	Strong for education	Strong for deep reasoning	Good for data analysis & visual tasks

Recommendations for Forward-Thinking Institutions and Individuals

To remain competitive and truly AI-empowered, academic institutions and individuals must shift their strategies. Here’s how:

Immediate Steps:

• Audit All Custom GPTs: Identify which GPTs are still running your requests and for what purpose.
  

• Rebuild in ChatGPT Projects: Port key GPTs to the Projects interface using GPT-4o or newer.
  

• Train Faculty & Designers: Invest in skill development for using Projects’ modular features, memory, and “More Models” access.
  

Infrastructure Adjustments:

• Expand Model Access: Use “More Models” in ChatGPT to experiment with O-series and GPT-4.1.
  

• Adopt Memory Wisely: Leverage Projects’ memory for tutors, coaches, and agents supporting longitudinal learning journeys.
  

• Embed AI Into LIPS: Align Projects with LIPS (Life-Interest-Project-System) and CARE to build cognitive scaffolding and personal growth tools within the University 365 environment.
  

GPTs vs Projects A Matter of Budget ?

Access Across Subscription Plans

Feature	Free Plan	Plus Plan	Team Plan	Enterprise Plan
Use GPTs	✅	✅	✅	✅
Create GPTs	❌	✅	✅	✅
Access Projects	❌	✅	✅	✅
Share Projects	❌	❌	❌	❌

• Free Plan:
  

  • Use GPTs: Free users can access and use GPTs from the GPT Store.

  • Create GPTs: Creation of Custom GPTs is not available.

  • Access Projects: Projects are not accessible.
    

• Plus Plan:
  

  • Create GPTs: Users can create and customize their own GPTs.

  • Access Projects: Full access to Projects for personal organization.

  • Share Projects: Sharing Projects is not supported.
    

• Team & Enterprise Plans:
  

  • Enhanced Collaboration: While GPTs can be shared among team members, Projects remain individual.

  • Administrative Controls: Additional features like user management and centralized billing are available.

What About Competition? What Google Gemini or Athropic Claude Offer ?

In the evolving landscape of AI tools, OpenAI's Custom GPTs and ChatGPT Projects have set benchmarks for customization and project management. However, other major players like Google and Anthropic have introduced their own equivalents, each with unique features and limitations.

Google Gemini: Gems and Project Astra

Gems: Google's answer to OpenAI's Custom GPTs, Gems are customizable AI chatbots that users can tailor with specific instructions and data. They integrate seamlessly with Google services like Gmail, Google Drive, and YouTube, enhancing their utility across various contexts. Gems are shareable and can be published for broader access, similar to OpenAI's GPT Store.

Project Astra: While not a direct equivalent to ChatGPT Projects, Project Astra represents Google's initiative towards more advanced, agentic AI experiences. It focuses on integrating AI capabilities across Google's ecosystem, enabling more complex, multi-step tasks.

Model Limitations: Gems utilize the Gemini Advanced model, which currently is Gemini 1.5 Pro. Users cannot select different models for their Gems, and access to creating Gems requires a subscription to Gemini Advanced.

Anthropic Claude: Projects

Projects: Anthropic's Claude offers 'Projects' as a feature for organizing chats and knowledge. These are akin to OpenAI's ChatGPT Projects, allowing users to structure their interactions and data. However, unlike OpenAI's Custom GPTs, Claude's Projects are not designed for public sharing or publishing.

Model Limitations: Claude's Projects leverage the Claude 3.5 Sonnet model, providing a substantial context window of up to 200,000 tokens. This allows for handling extensive data within a single project. 

Comparative Overview OpenAI vs Anthropic vs Google

Feature	OpenAI Custom GPTs	OpenAI Projects	Google Gems	Google Project Astra	Claude Projects
Shareability	✅ Yes	❌ No	✅ Yes	❌ No	❌ No
Model Selection	❌ Fixed (GPT-4-turbo)	✅ Yes	❌ Fixed (Gemini 1.5 Pro)	❌ No	❌ Fixed (Claude 3.5 Sonnet)
Subscription Required	✅ Yes	✅ Yes	✅ Yes	✅ Yes	✅ Yes
Integration with Services	✅ Limited	✅ Yes	✅ Extensive (Google Services)	✅ Yes	✅ Yes
Context Window	Up to 128k tokens	Up to 128k tokens	Up to 1M tokens	Up to 2M tokens	Up to 200k tokens

Conclusion: Embrace the Future Before It Leaves You Behind

While OpenAI's Custom GPTs and Projects offer a blend of shareability and advanced project management, Google's Gems provide a shareable, service-integrated alternative, albeit with fixed model usage. Anthropic's Claude Projects focus on structured organization without public sharing capabilities. Like OpenAI with ChatGPT "GPTs" and "Projects" features, each platform presents unique advantages and limitations, and the choice between them should align with specific user needs regarding collaboration, customization, and integration. The future of applied AI demands adaptive, memory-enabled, multimodal reasoning agents, not static, turbocharged chatbots stuck in 2023. OpenAI’s Custom GPTs were a powerful start, but, fixed to GPT-4-turbo, they are no longer sufficient. This also applies to Gemini Gems (fixed to Gemini 1.5 Pro) and Claude Projects (fixed to Claude 3.5 Sonnet), which is sufficient for most cases.

If you'r a OpenAI ChatGPT user, by transitioning to ChatGPT Projects, instead of GPTs, you unlock a new generation of intelligent agents,capable of designing, resolving, teaching, mentoring, and evolving.

Let this moment be a turning point. Your strategic decisions today will define whether you lead or lag in the AI-powered revolution.