What to expect from GPT-5
TL;DR
GPT-5 is expected to launch in July 2025, but it's not out yet. Early information suggests it will improve memory, reasoning, and multimodal capabilities significantly. However, it’s unlikely to unlock entirely new use cases. More likely, it will refine and streamline existing ones.
Should you prepare to act?
If you're already using GPT-4 effectively: wait and evaluate once GPT-5 is released.
If you’re not using generative AI in your processes yet: now is the time to start. Whether GPT-5 is a big leap or a small step, waiting longer just leaves you behind.
GPT-5 is not released yet, but it's already shaping discussions across the AI world. With growing anticipation and early leaks surfacing from insiders, many businesses are wondering: what will it offer, how different will it be from GPT-4 (turbo etc), and should we start preparing now? In this post, we take a critical, practical look at what GPT-5 might mean for your organization.
What Is GPT-5 and Why It Matters
GPT-5 is expected to launch in July 2025, but as of now, it remains unreleased. Still, the features it's rumored to include, such as unified multimodal inputs/outputs, longer memory, better reasoning, and more agent-like behavior, are generating both excitement and skepticism.
What makes GPT-5 notable is its ambition to consolidate tasks that previously required several tools or workflows. It aims to become a more integrated, context-aware assistant capable of understanding and acting across a variety of data types (text, image, audio). For businesses, this sounds promising, but is it essential on day one?
The short answer: it depends. If you're already using GPT-4 effectively (e.g. for internal assistants or customer support chatbots), you may not need to rush into GPT-5. The gains will likely be in efficiency and versatility, not necessarily in entirely new capabilities.
That said, the shift toward agentic behaviors and memory-driven interactions could unlock new possibilities - especially for those working with complex, multi-step processes. The real business value will depend on implementation, not just model release.
Key Advancements and Capabilities
Technical Improvements
You are probably wondering: “What are the changes this time?”. Well, here are some things you can expect:
Massively expanded context windows, possibly up to 2 million tokens. Here, the context window refers to the total amount of text a language model can consider at once. Both what you input and what it outputs. The larger the window, the more information the model can work with in a single interaction. Now, a window of 2 million tokens [SA1] [JV2] could let the model handle entire books, full project archives, or months-long conversations. But here's the catch: historically, large context windows often lead to performance trade-offs, such as slower responses or less relevant answers. It will be important to test whether GPT-5 can maintain coherence at these lengths. Also, more tokens = more cost. So should we all start pumping massive books into our prompts? Probably not: optimizing for both performance and cost will remain crucial.
Long-term memory: Instead of resetting every session, GPT-5 may remember key details about users, tone, preferences, and context across conversations. For businesses, this could be useful in applications like CRM automation, sales follow-up, or even personalized HR tools. But adoption will boil down to whether memory is secure, explainable, and manageable.
Unified multimodal interface: GPT-5 is expected to support inputs and outputs across text, image, audio, and possibly video. All within a single prompt. That simplifies workflows and removes the need to switch between tools. But whether that works reliably out-of-the-box remains to be seen.
New Use Cases Or Refining Existing Ones?
While GPT-5 brings better performance and usability, and most of the commonly mentioned use cases (code generation, summarization, legal review, creative work, customer service, etc.) are not new. What's changing is that these tasks may become faster, more integrated, and less error-prone thanks to better reasoning, memory, and context handling.
Beyond Chat: From Tools to Agents
One of GPT-5’s most hyped features is its agentic behavior. It’s supposed to go beyond replying to prompts and begin executing tasks: booking meetings, filing reports, managing apps, and so on.
But let's be clear: these abilities aren’t off-the-shelf. Companies will still need to connect GPT-5 to their internal systems, build secure workflows, and define how far the AI can act without oversight. GPT-5 may provide the foundation, but deploying real agents in production will require engineering, testing, and monitoring.
Costs, Access, and Implementation
Pricing
Exact pricing isn’t public yet. GPT-5 will likely follow a tiered model similar to GPT-4, with enterprise access offered via Azure OpenAI or the OpenAI API. Expect higher rates for longer context windows and memory features, especially for multimodal or always-on agents.
Training and Architecture
Synthetic data is playing a growing role, especially in domains like math and coding, where answers can be automatically verified. This could lead to sharper performance in practical, enterprise-relevant tasks.
GPT-5 is also expected to use Mixture-of-Experts (MoE) techniques, improving efficiency by routing different tasks to specialized sub-networks.
Trust, Compliance, and Risk
Security, Privacy, and Compliance
As GPT-5 introduces persistent memory and more integrated actions, compliance risks grow. For businesses in Europe, this raises an obvious question: Will the model be compliant with GDPR and local data regulations?
OpenAI has suggested that users will be able to view, edit, or delete what the model remembers. But enterprise deployment will still require governance mechanisms, such as:
Defining what memory can store
Isolating sensitive data from broader training loops
Ensuring audit trails for actions taken by agents
Hosting GPT-5 in Azure environments will likely remain the preferred route for enterprises as Microsoft doesn’t share any of their data with OpenAI. Additionally, this route gives you as a business enhanced control, identity management, and data residency guarantees.
Risks, Limitations, and Responsible AI
Hallucination rates are reportedly below 15%, which is promising. But:
GPT-5 will still require human oversight in high-stakes contexts
Explainability remains limited: it’s better at reasoning, but still a black box
Agent autonomy introduces risk: Who’s responsible when a workflow breaks or an email is misfired?
Ultimately, GPT-5 is a more capable system, but the fundamentals of responsible AI design, monitoring, and testing still apply.
Final Take: Should You Act Now?
We don’t know yet. GPT-5 isn't released, and much of what’s known is speculative. But once it’s out and independently tested, we’ll get a better sense of whether it’s truly groundbreaking.
For now, if you're already using GPT 4, you might start evaluating how GPT-5 could enhance or streamline your existing workflows, but only once it's live and validated.
If you haven't adopted generative AI at all yet, there’s no more time to lose. The fundamentals won’t change with GPT-5: efficiency gains, automation potential, and competitive pressure will only increase.
In the meantime, if you have any questions about Generative AI, feel free to reach out.
Author: Joran Vergauwen
Joran Vergauwen
Hi, I’m Joran. I’m the AI Lead at Plainsight and one of the cofounders of the Generative AI Belgium meetup group, a vibrant community of over 3,000 AI enthusiasts. I have over five years of hands-on experience in AI, including time spent building my own GenAI startup. On that journey, I’ve learned that chasing the latest tech trends isn’t enough though. The real value comes from focusing on real problems and solving them effectively.
I’m always experimenting, trying out new tools, and exploring what’s possible with AI. My background has taught me to stay curious and practical, balancing innovation with impact.
When I’m not deep in code or strategy, you’ll find me reading, running, or playing the piano. Want to connect or chat about AI? Feel free to reach out!
Anxious to know what Plainsight could mean for you?
Read more “About us”
Consider “your career at Plainsight 🚀”
Any questions? “Contact us”