The release of GPT-5 has ignited a wave of discussions, as the latest iteration from OpenAI prompts both excitement and scrutiny.
GPT-5 sees OpenAI delivering a more consumer facing, broad use, AI "system" that indicates at least some of their focus is on realising the dream of developing a ubiquitous AI assistant, all whilst they delivered only incremental improvements to the underlying model's intelligence.
GPT-5 combines all the different model types (deep thinking, fast response, coding, etc.) into one "system" with a real-time "router" that picks the relevant model under the hood to use for each response - Read More
This is an attempt to help the everyman get more out of ChatGPT and lower the learning curve to get value from the "product", which they need to do if they want to realise mass adoption and mass paid subscriptions that will help reduce their current unsustainable operating deficits.
It offers a much larger context window, reduced hallucinations, new “personalities,” and gets chart topping performance scores across almost all benchmarks, especially coding - Read More
While performance is best-in-class, it feels more like a GPT-4 upgrade than a leap towards AGI, but Altman's PR shenanigans to raise hype/money meant GPT-5 was always going to fall short of expectation - Read More
The GPT-5 launch was messy, with staggered announcements, buggy live demos, and a rollout that felt rushed - Read More
To their credit, OpenAI responded to the frustrations of power users and reinstated the model picker and also brought back legacy models like GPT-4o after everyday users complained about personality changes meaning it felt like they "lost a friend" - Read More
AI nerds saw the release as underwhelming compared to the hype, but most mainstream users are likely to see this as an incredibly substantial upgrade.
GPT-5 is cheap and API pricing is going to make high-level reasoning models more affordable than ever - Read More
The pricing strategy hints at an aggressive push for adoption and perhaps an attempt to dethrone Anthropic's Claude as the best AI tool for developers - which has seen Anthropic get significant market share through a winning price and capability combination.
It will be interesting to see if this causes a ripple effect on pricing from other providers in order to try to stave off loss of market share.
Looking forward, it's also interesting to see where OpenAI's future is heading as a company, and it may not be what you're expecting - Read more
In summary, GPT-5 is impressive as an indication of what's to come, but doesn't replicate the giant leaps forward we saw when moving from GPT-3 to GPT-4. There is a lot of time and effort that has been invested in making GPT-5 easier to use and easier to get tangible value from for everyday users. It is smarter, a lot more capable and can do "things" for users much more autonomously and with less guidance - which will be great for the average Joe. On the other hand, it doesn't feel like it gets us closer to AGI or superintelligence. All in all this feels like commercial realities were factored into the product roadmap for GPT-5.
Curious how your business can get started with AI? Talk to our team about our Kickstart program, designed to introduce, educate, and integrate AI into your team.