Can OpenAI pull ahead again now that Google has caught up on AI?

Can OpenAI pull ahead again now that Google has caught up on AI?

OpenAI has released a new flagship model, GPT-5.2, and is pitching it as its strongest system so far for professional, real-world work. The company says it performs better on widely used benchmarks for coding, mathematics and science, and it is positioning the model as an upgrade for serious users rather than a flashy demo, the New York Times reported.

The timing is not accidental. Just before Thanksgiving, Google publicly boasted that its latest model, Gemini 3, had overtaken OpenAI. OpenAI’s response arrives less than a month later, and the message is simple: the company that kicked off the consumer AI boom still intends to define the cutting edge.

For a couple of years after ChatGPT’s launch in late 2022, OpenAI had a clear lead in what the industry calls “foundational models”, the general-purpose systems that power chatbots, coding tools and enterprise assistants. That advantage is now much harder to see.

Over the last year, rival labs in the United States and China have released models that are comparable on many tasks, and sometimes stronger in specific areas. In the same window, Anthropic has launched a new Claude model positioned as peer-level, and Runway has claimed gains that exceed OpenAI in video generation benchmarks relative to Sora. The direction of travel is clear even if the rankings shift week to week: this is no longer a one-company race.

Several people who track model performance describe the field as converging. The core recipe for building and improving these systems is now widely understood across top labs, which means “surprise” leaps are harder to sustain.

What makes this moment more sensitive for OpenAI is the business math. Sam Altman has suggested the company expects revenue to reach a run rate equivalent to about $20 billion annually by the end of 2025. That is huge growth by any normal standard, but it does not solve OpenAI’s larger problem: it is still far from profitability.

The company has also signalled extraordinary spending commitments for computing power over the coming years. That is not just research expenditure; it is the cost of training, serving, and constantly upgrading models at global scale. As rivals close the technology gap, OpenAI’s financial gap becomes harder to ignore.

Inside OpenAI, the consumer chatbot is still the centre of gravity. The company reportedly has around 800 million weekly ChatGPT users, and a relatively small fraction paying for subscriptions supplies a major share of revenue. That makes ChatGPT both an asset and a vulnerability: if users drift elsewhere because rivals feel “good enough”, OpenAI’s leverage with investors and enterprise customers weakens.

After Google’s Gemini 3 splash, Altman reportedly sent a “code red” memo urging staff to prioritise upgrades to ChatGPT. The focus was practical: make it faster, reduce unnecessary refusals, and improve personalisation. In other words, win on product experience, not just model scores.

OpenAI is also moving pricing upward. The company has indicated that GPT-5.2 will cost materially more than earlier models, reflecting both improved capability and the rising cost of compute. This is the harder part of the new competition: charging more at the same time rivals offer similar performance.

On the business side, OpenAI is pushing deeper into enterprise software, arguing that AI will reshape nearly every category of workplace tools. But it has an obvious handicap when competing with a company like Google, which can bundle AI into established products like Gmail and Google Docs. OpenAI is trying to counter this by building its own distribution channels, including a newly introduced web browser designed to work tightly with its AI.

OpenAI is also exploring new ways to monetise free users. That includes advertising experiments and more direct commerce, where ChatGPT could steer users to products or services and take a cut of the transaction. It is a meaningful shift, especially for early employees who prized the idea that OpenAI would not become another ad-driven platform.

The problem is structural: chat interfaces do not naturally fit traditional ads the way search results do. So OpenAI’s challenge is not merely “add ads”, but “invent a format that users tolerate and regulators accept”.

GPT-5.2 may be a technical step forward, especially in coding and specialised professional work. But the more revealing signal is what surrounds the release: tighter competition, heavier spending, and a fight across many fronts, not just model quality.

The next phase for OpenAI will likely be decided by product speed, distribution, partnerships, and business discipline as much as by raw intelligence. In a market where the top systems are starting to feel interchangeable to ordinary users, that may be the toughest shift of all.