On Thursday, OpenAI announced far more than just another model release. Alongside GPT-5.6 came a series of changes that signal a broader shift in strategy and where the company believes AI is heading next. If I didn’t know better, I’d say they’ve been taking notes from Anthropic’s playbook (😜), because many of these moves closely resemble the direction Anthropic has been pursuing for some time.
Have We Reached Peak Consumer AI?
The first thing to say is that GPT-5.6 is not a consumer model. I’d argue it’s not even really a prosumer model. This is a model built for people who spend their working day using LLMs, and OpenAI’s own marketing reflects that.
This raises an interesting question: is consumer-facing AI now as capable as it needs to be?
For the vast majority of people, AI is used for discovery, research, learning, summarising information, drafting emails and supporting decisions. Do those users really need significantly more capable models? I’d argue the answer is probably no.
Today’s frontier models are already more than capable enough for most consumer use cases. Of course, they still need to become faster, cheaper and available to everyone (not just the 5% that are paid subscribers) but I’m no longer convinced that continually increasing model intelligence will materially improve the experience for the average user. We may have reached the point where models like GPT-5.5 are already “good enough” for most consumers, leaving the next wave of AI innovation to focus elsewhere.
The second thing that struck me was just how “Anthropic” this launch felt. OpenAI has traditionally released individual flagship models, but GPT-5.6 arrives as a family: Sol (large), Terra (medium) and Luna (small). Even the naming convention looks familiar, mimicking Anthropic’s Opus, Sonnet and Haiku. It’s a small detail, but it reinforces the sense that OpenAI is borrowing more than a little from Anthropic’s product strategy.
The third thing that stood out to me is just how enterprise-focused this launch feels. OpenAI’s previous launches have always had a strong consumer story - faster conversations, better image generation or more natural interactions. GPT-5.6 is different. Almost every benchmark, demo and example centres on coding, research, cybersecurity, scientific reasoning or knowledge work. OpenAI is prioritising people who create value with AI, rather than people who simply use AI.
There’s some good commentary from Simon Willison, as always, if you want to dive a little deeper.
OpenAI’s Enterprise Pivot
If you needed any more convincing that OpenAI is pivoting towards Anthropic’s playbook, look no further than ChatGPT Work. The name alone echoes Claude Cowork, and the similarities don’t stop there.
This is not two companies independently arriving at the same idea. Anthropic has spent the past few years concentrating on developers, knowledge workers and enterprise customers, while OpenAI focused primarily on building the world’s largest consumer AI product. Anthropic’s strategy has been paying off over the last six months, and OpenAI are now pivoting away from consumers.
First it responded to Claude Code with Codex. Now it is responding to Claude Cowork with ChatGPT Work. After three and a half years of prioritising consumers, OpenAI is moving its attention towards the professional and enterprise users who are prepared to pay considerably more for AI.
Reuters described ChatGPT Work as OpenAI’s answer to Anthropic’s enterprise push, reflecting the increasingly fierce battle for enterprise AI customers. That’s exactly what this feels like as OpenAI reposition ChatGPT for the workplace.
So what is ChatGPT Work? Think of it as ChatGPT evolving from an assistant that answers questions into one that executes work. It can gather context from your files and connected applications, plan and carry out multi-step tasks, and produce finished documents, spreadsheets, presentations and websites. This is OpenAI betting that the future of AI is more than conversations, it’s helping people get stuff done.
OpenAI’s Super App Gamble
Having spent the last 24 hours using GPT-5.6 and living inside the new ChatGPT apps, I’ve come away with one overriding conclusion: I’m not convinced that any of this is good for consumers.
OpenAI has been very clear about where it’s heading. ChatGPT is no longer a conversational AI. It’s becoming a super app for knowledge work. A place to code, build presentations, analyse data, carry out research, connect to your tools and increasingly delegate work to AI agents.
That sounds great in theory, but the experience feels very different in practice.
One of the biggest changes yesterday was the switch from the Codex app to the new ChatGPT app, while the original ChatGPT was quietly renamed ‘ChatGPT Classic’. Yet despite all of those changes, I’ve ended up exactly where I was before: using two separate apps. That’s disappointing because it suggests OpenAI hasn’t unified the experience, it’s just (confusingly) rebranded it.
When I want to build something, write code or complete a complex task, I naturally reach for the new ChatGPT app (ex. Codex). But when I simply want to ask a question, explore an idea or have a conversation, I keep going back to ChatGPT Classic. Chats have become something of a third-class citizen in the new ChatGPT app experience, and for me that fundamentally changes what ChatGPT feels like.
The “Classic” label also feels significant. In software, “Classic” rarely means “this is the future”. It usually means “this is here for now”. Whether or not OpenAI intends to retire the app, it’s clear that the new ChatGPT (ex. Codex) app is the product they want everyone to use.
Maybe that’s exactly what OpenAI wants. If its future lies in enterprise customers rather than consumers, then turning ChatGPT into the operating system for knowledge work makes complete commercial sense. But it’s also a gamble. In trying to become everything for everyone, OpenAI risks making the world’s biggest consumer AI platform less usable for the people who made it successful in the first place.
OpenAI’s big gamble
I don’t think the biggest winner from yesterday’s announcements was OpenAI. It was Google.
Over the past two years, Gemini has established itself as the number two consumer AI platform behind ChatGPT.. At the same time, Google has been steadily embedding AI across Search, Gmail, Chrome, Android, YouTube, Maps and Workspace. Unlike OpenAI or Anthropic, it doesn’t need Gemini to become a multi-billion dollar standalone business. AI simply makes its existing ecosystem more valuable.
While OpenAI and Anthropic are increasingly competing for developers, knowledge workers and enterprise customers, Google has the luxury of continuing to optimise for billions of everyday consumers.
That’s because OpenAI is making a big bet.
It’s betting that the future of AI isn’t consumer subscriptions or paid advertising. It’s betting that the real commercial opportunity lies with enterprise customers. Consumer subscriptions haven’t scaled, with the overwhelming majority of ChatGPT users still choosing not to pay. At the same time, OpenAI’s paid advertising products remain in their infancy and are still far from becoming a major source of revenue.
Enterprise is different. It offers larger contracts, higher-value customers and significantly more revenue per account.
But every big bet comes with a trade-off. In chasing Anthropic, OpenAI is also gambling with its lead in consumer AI. If ChatGPT continues to evolve into the operating system for knowledge work, it leaves Google with a clear opportunity to become the dominant AI platform for everyday consumers.
The irony is hard to ignore. In trying to win the most valuable customers, OpenAI may be making it easier for Google to win the biggest market.
Other Tidbits
“The future is already here, it’s just not evenly distributed.“
William Gibson