OpenAI has unveiled GPT-5.4, its latest flagship model, and the timing reveals something important about how the company sees its path forward. This is not a model designed for casual conversation. It is squarely aimed at knowledge workers who spend their days wrestling with spreadsheets, financial models, legal documents, and presentations.
GPT-5.4 is billed as OpenAI's "most capable and efficient frontier model for professional work." In tests of knowledge work across 44 occupations, the model matched or exceeded industry professionals in 83 per cent of comparisons. Those benchmarks included real-world tasks such as generating sales presentations, accounting spreadsheets, and manufacturing diagrams.
The business logic is straightforward: general-purpose chatbots are increasingly commoditised. Every tech company has one now. But specialised tools that can actually help accountants build financial models or lawyers analyse contracts faster? That is where margin lives. OpenAI's chief financial officer Sarah Friar recently told CNBC that the company expects enterprise customers to grow from 40 per cent of its business to 50 per cent by year's end. GPT-5.4 is the sharp end of that strategy.
Making the model smarter
GPT-5.4 is OpenAI's most factual model yet; it is 33 per cent less likely to be false in individual claims and 18 per cent less likely to contain any errors in full responses compared to GPT-5.2. For anyone who has watched AI models hallucinate their way through important documents, that matters. A spreadsheet error buried in 50 rows of calculations can be expensive. A false legal citation in a contract review is worse.
The model also gains native computer-use capabilities, a first for OpenAI's mainstream product line. GPT-5.4 is the company's first general-purpose model with native computer-use capabilities, excellent at writing code to operate computers via libraries like Playwright and issuing mouse and keyboard commands. This means the model can do more than just generate code; it can actually execute workflows across multiple applications.
Pricing reflects the premium positioning. OpenAI is charging $2.50 per million input tokens for GPT-5.4, up from $1.75 for GPT-5.2. That is a 43 per cent increase. Developers will feel that in their cloud bills.
The financial squeeze
None of this happens in a vacuum. OpenAI generated $13.1 billion in revenue in 2025, up from its $10 billion target, while burning through $8 billion, lower than its $9 billion target. Those figures sound healthy until you realise the company is still unprofitable and deeply dependent on continued investor funding.
The company previously touted $1.4 trillion in infrastructure commitments but is now telling investors it plans to spend $600 billion by 2030. Even the reduced number is enormous. OpenAI is projecting total revenue of more than $280 billion by 2030, with nearly equal contributions from consumer and enterprise businesses. That would require explosive growth, and there is no guarantee it will arrive.
A pragmatic compromise
The shift toward professional tools represents a sober choice. Consumer chatbots are useful but hard to monetise in ways that justify billion-dollar infrastructure budgets. Enterprise customers, by contrast, will pay premium prices for tools that save money or time on real workflows. A company that uses GPT-5.4 to speed up contract review or financial modelling sees a clear return on investment.
That does not mean OpenAI has solved its fundamental problem. The company is still burning through cash faster than it is earning revenue. The company has faced mounting concerns about whether it can ever generate enough revenue to cover its costs. GPT-5.4 is a bet that enterprise software is the answer, and it may well be. But bets are not the same as certainties, and investors are watching closely.
For now, OpenAI is doing what any rational business would do in this position: double down on the revenue streams that work best, improve the product quality relentlessly, and hope the growth curve stays steep enough to justify the spending. Whether that strategy holds up depends partly on how much better GPT-5.4 actually performs on real-world professional tasks, and partly on whether the macroeconomic wind continues to blow in AI's favour.