GPT-5 has landed, marking the largest AI model upgrade in history. The question worth asking: is this the moment where AI becomes part of everyday life for everyone, or the stage where improvements feel smaller to those already working at the frontier?
In other words, are we closing the gap between beginners and experts, or cementing the divide?
For power users, GPT-5 brings welcome upgrades. It is faster, sharper, more reliable, and less prone to hallucinations (When AI makes up facts or gives false info). But if you already had prompt libraries, workflow hacks, and model chaining habits in place, many of the problems GPT-5 solves were already under control. The ceiling has moved, but not so far that it feels like a revolution. The divide might not be closing as a result. In fact, some might argue it is giving experts even more tools to pull further ahead.
For beginners, it is a different story. GPT-5 removes much of the friction that used to kill AI adoption early. No more juggling models, reducing unpredictable quality, or writing off outputs you could not trust. For most free users there is now one model, the best model, and it just works. The floor is so much higher that your very first conversation feels capable and confident. Yes, there are still hallucinations and moments of agreement when the AI should push back. But the baseline is now so solid that more people will stick around long enough to see what AI can really do for them.
The question for anyone watching this shift is whether this higher floor will genuinely pull beginners closer to power users, or whether the experts will keep stretching the ceiling faster than newcomers can catch up. That is where the gap teeters on a knife edge for NZ in particular.
Explaining the Divide
Three ideas help explain why experienced users and beginners will see GPT-5 so differently:
Product Lifecycle Theory: AI is hitting the flattening part of the S curve. The early jumps from GPT 3 to GPT 4 felt like revolutions. Now the improvements are more incremental for those already at the edge, just like the iPhone shifting from Retina Display level leaps to slightly better cameras.
Diminishing Marginal Utility: Each improvement feels smaller once you have already been getting high value. For power users, halving hallucinations or shaving two seconds off a response is welcome but not transformative. For beginners, those changes still feel like a leap.
The Matthew Effect: The rich get richer principle. Power users will take GPT-5’s extended reasoning, advanced voice, and deeper integrations and compound their advantage. Beginners get a better starting point, but the ceiling still rises for the experts.
Why New Zealand should care
For NZ businesses, whether GPT-5 closes the gap or cements the divide will come down to how quickly these raised-floor capabilities are put to use. In a market where many businesses are still in the testing phase with AI, GPT-5’s raised floor is a gift.
A small business in Hawke’s Bay can now access enterprise grade reasoning without paying for Pro.
A Central Otago tourism operator can draft personalised itineraries without hitting formatting or accuracy walls.
An educator can build teaching resources without fighting the tool.
For those already at the frontier, GPT-5 is about leverage: plugging it into existing workflows, talking to improved custom GPTs connected to Google Drive, and shipping ideas faster than ever.
For those who thought AI was all hype or too messy to bother with, this could be the tipping point that moves them from dabbling to real adoption.
The Great AI divide in Aotearoa
But while GPT-5’s potential is huge, the benefits will not be felt evenly across Aotearoa. Long-standing gaps in access, skills, and trust mean some communities are better placed to take advantage of this moment than others.
The AI Forum’s State of AI in New Zealand report revealed a stark reality: only a small minority of organisations are AI mature, while most are still experimenting or not considering AI at all. Many have no plan, no strategy, and no urgency.
That gap is also not evenly spread. Pasifika tech leaders like Namulau'ulu Nu’uali’i Eteroa Lafaele have warned that without deliberate inclusion, entire communities could be “coded out of the future". 25% of Pacific households still lack reliable devices and connectivity. Even the best AI model in history is meaningless if people cannot access or confidently use it.
Digital inequity and lack of trust remain major barriers. Communities that have been let down by past waves of technology adoption are understandably cautious. If GPT-5 is going to close the divide, it must be paired with targeted efforts to build skills, representation, and community confidence.
GPT-5’s raised floor could be the tipping point that moves thousands of Kiwi organisations out of trial mode and into serious adoption. That will only happen if access is matched with capability, and capability is distributed fairly.
GPT-5 Feature Review
So what exactly does this raised floor look like in practice? GPT-5 brings a set of upgrades that change both the experience for everyday users and the possibilities for advanced ones.
Release and availability
OpenAI has called this a huge day: GPT-5 is now live for 700 million weekly active users
It is free for everyone and replaces GPT 4o and GPT 4.5 entirely. Huge benefit for most, but the reduced access to more models will frustrate many
The GPT-5 model is rolling out across all ChatGPT accounts, Microsoft CoPilot, Cursor, Lovable and more
Core capabilities
Performance: Described as best in class performance with agentic coding research abilities and significantly fewer hallucinations. Funny enough, it crashed multiple times on me yesterday and failed to respond to basic outputs. The fix seemed to be in last night!
Speed: Much faster than GPT 4o, with many users saying they are 5 to 10 times more productive. Can vouch for this, much quicker.
Longer context handling: novels, 10,000 line codebases, medical histories, financial reports, without losing track. This is the biggest uplift for Power Users in my opinion. The 'drift' in thinking taking place previously was the biggest frustration in previous models.
Coding: The top coding model according to benchmarks, narrowly beating Anthropic’s Claude Opus 4.1. The vibe coding capabilities in GPT-5 are a welcome boost, helping OpenAI secure further user lock in. Transparently, I couldn't recreate 50% of what the X/Twitter or demo users were able to recreate.
Benchmarks and reliability
Hallucinations in thinking mode down from 13 percent to 1.6 percent
General factual error rate 45% lower than GPT 4o
Incorrect agreement rates down from 14 percent to under 6 percent
Enterprise and API upgrades
The largest enterprise AI upgrade in history
Hundreds of millions of Microsoft CoPilot and Teams users now on GPT-5
API wrappers (software layers that package AI models into usable tools) and software systems swapping GPT 4o for GPT-5 behind the scenes, meaning customer service bots, productivity tools, and other systems just get sharper and faster
Vibe coding explosion
GPT-5 makes it easy to build apps, tools, and dashboards from plain language prompts
Canvas mode lets you run and render code as if it were a built in web browser
Cursor has made GPT-5 its default model, calling it the smartest coding model they have ever tried
Expect more everyday users creating products without needing a developer
Other notable features
Advanced Voice Mode in CustomGPTs: faster, more human, and can act as a personal strategist with full Google Drive access
Multiple operational modes including auto, thinking, and pro for heavy lifting tasks
Wrapping up
The floor has risen. The tools are in place. The capabilities are in everyone’s hands. Whether it closes the gap between beginners and power users or cements it will depend on how quickly access turns into skill.
Written by Mike ✌
Passionate about all things AI, emerging tech and start-ups, Mike is the Founder of The AI Corner.
Subscribe to The AI Corner
