Sam Altman turned OpenAI into a cultural phenomenon with ChatGPT.
Now, three years later, he’s chasing where the real money is: Enterprise.
Last week’s rollout of GPT-5, OpenAI’s newest artificial intelligence model, was rocky. Critics bashed its less-intuitive feel, ultimately leading the company to restore its legacy GPT-4 to paying chatbot customers.
But GPT-5 isn’t about the consumer. It’s OpenAI’s effort to crack the enterprise market, where rival Anthropic has enjoyed a head start.
One week in, and startups like Cursor, Vercel, and Factory say they’ve already made GPT-5 the default model in certain key products and tools, touting its faster setup, better results on complex tasks, and a lower price.
Some companies said GPT-5 now matches or beats Claude on code and interface design, a space Anthropic once dominated.
Box, another enterprise customer, has been testing GPT-5 on long, logic-heavy documents. CEO Aaron Levie told CNBC the model is a “breakthrough,” saying it performs with a level of reasoning that prior systems couldn’t match.
Behind the scenes, OpenAI has built out its own enterprise sales team — more than 500 people under COO Brad Lightcap — operating independently of Microsoft, which has been the startup’s lead investor and key cloud partner. Customers can access GPT models through Microsoft Azure or go directly to OpenAI, which controls the API and product experience.
Still, the economics are brutal. The models are expensive to run, and both OpenAI and Anthropic are spending big to lock in customers, with OpenAI on track to burn $8 billion this year.
Winning over enterprise
GPT-5 is significantly cheaper than Anthropic’s top-end Claude Opus 4.1 — by a factor of seven and a half, in some cases — but OpenAI is spending huge amounts on infrastructure to sustain that edge.
For OpenAI, it’s a push to win customers now, get them locked in and build a real business on the back of that loyalty.
Cursor, still a major Anthropic customer, is now steering new users to OpenAI. The company’s co-founder and CEO Michael Truell underscored the change during OpenAI’s launch livestream, describing GPT-5 as “the smartest coding model we’ve ever tried.”
Truell said the change applies only to new sign-ups, as existing Cursor customers will continue using Anthropic as their default model. Cursor maintains a committed-revenue contract with Anthropic, which has built its business on dominating the enterprise layer.
As of June, enterprise makes up about 80% of its revenue, with annualized revenue growing 17x year-over-year, said a person familiar with the matter who requested anonymity in order to discuss company data. The company added $3 billion in revenue in just the past six months — including $1 billion in June alone — and has already signed triple the number of eight- and nine-figure deals this year compared to all of 2024, the person said.
Anthropic said its enterprise footprint extends far beyond tech.
Claude powers tools for Amazon Prime, Alexa, and AIG, and is used by top players in pharma, retail, aviation, and professional services. The company is embedded across Amazon Web Services, GCP, Snowflake, Databricks, and Palantir — and its deals tend to expand fast.
Average customer spend has grown more than fivefold over the past year, with over half of business clients now using multiple Claude products, the person said.
Excluding its two largest customers, revenue for the rest of the business has grown more than elevenfold year-over-year, the person said.
Even with that broad reach, OpenAI is gaining ground with enterprise customers.
GPT-5 API usage has surged since launch, with the model now processing more than twice as much coding and agent-building work, and reasoning use cases jumping more than eightfold, said a person familiar with the matter who requested anonymity in order to discuss company data.
Enterprise demand is rising sharply, particularly for planning and multi-step reasoning tasks.
GPT-5’s improvement
GPT-5’s traction over the past week shows how quickly loyalties can shift when performance and price tip in OpenAI’s favor.
AI-powered coding platform Qodo recently tested GPT-5 against top-tier models including Gemini 2.5, Claude Sonnet 4, and Grok 4, and said in a blog post that it led in catching coding mistakes.
The model was often the only one to catch critical issues, such as security bugs or broken code, suggesting clean, focused fixes and skipping over code that didn’t need changing, the company said. Weaknesses included occasional false positives and some redundancy.
Vercel, a cloud platform for web applications, has made GPT-5 the default in its new open-source “vibe coding” platform — a system that turns plain-English prompts into live, working apps. It also rolled GPT-5 into its in-dashboard Agent, where the company said it’s been especially good at juggling complex tasks and thinking through long instructions.
“While there was a lot of competition already in AI models, Claude was just owning this space. It was by far the best coding model. It was not even close,” said Malte Ubl, CTO of Vercel. “OpenAI was just not in the game.”
That changed with GPT-5.
“They at least caught up,” Ubl said. “They’re better at some stuff, they’re worse at other stuff.”
He said GPT-5 stood out for early-stage prototyping and product design, calling it more creative than Claude’s Sonnet.
“Traditionally, you have to optimize for the new model, and we saw really good results from the start,” he said about the ease of integration.
JetBrains has adopted GPT-5 as the default in its AI Assistant and in Kineto, a new no-code tool for building websites and apps, after finding it could generate simple, single-purpose tools more quickly from user prompts. Developer platform Factory said it collaborated closely with OpenAI to make GPT-5 the default for its tools.
“When it comes to getting a really good plan for implementing a complex coding solution, GPT-5 is a lot better,” said Matan Grinberg, CEO of Factory. “It’s a lot better at planning and having coherence over its plan over a long period of time.”
Grinberg added that GPT-5 integrates well with their multi-agent platform: “It just plays very nicely with a lot of these high-level details that we’re managing at the same time as the low-level implementation details.”
Pricing flexibility was a major factor in Factory’s decision to default to GPT-5, as well.
“Pricing is mostly what our end users care about,” said Grinberg, adding that cheaper inference now makes customers more comfortable experimenting. Instead of second-guessing whether a question is worth the cost, they can “shoot from the hip more readily” and explore ideas without hesitation.
Anton Osika, co-founder and CEO of Lovable, a company that builds an AI-powered tool that lets anyone create real software businesses without writing a single line of code, said his team was beta testing GPT-5 for weeks before it officially launched and was “super happy” with the improvement.
“What we found is that it’s more powerful. It’s smarter in many complex use cases,” Osika said, adding that the new model is “more prone to take actions and reflect on the action it takes” and “spends more time to make sure it really gets it right.”
Box‘s Levie said the biggest gains for him showed up in enterprise workflows that have nothing to do with writing code. His team has been testing the model for weeks on complex, real-world business data — from hundred-page lease agreements to product roadmaps — and found that it excelled at problems that tripped up earlier AI systems.
Levie added that for corporate use, where AI agents run in the background to execute tasks, those step-change improvements are critical, and can turn GPT-5 into a real breakthrough for work automation.
“GPT-5 has performed unbelievably well — certainly OpenAI’s best model — and in many of our tests it’s the best available,” he said.
— CNBC’s Kevin Schmidt contributed to this report.
WATCH: OpenAI launches GPT-5 model