This is not a typo: OpenAI may charge $20,000 a month for specialized artificial intelligence (AI) agents for “PhD-level” AI systems.
The company is also mulling AI agents for tailored applications such as a software developer for sorting and ranking sales leads and software engineering for $10,000 monthly. Another “high-income knowledge worker” agent would cost $2,000 per month, according to a report in The Information.
OpenAI investor SoftBank has committed $3 billion this year on the high-end agents, the news site said.
Though jarring to those who pay $20 monthly for low-end OpenAI services, such pricey models would help OpenAI produce revenue at a time when it is burning through cash. The company spent about $5 billion last year, The New York Times reported, and OpenAI is in talks to raise another $40 billion, CNBC reported.
“Certainly OpenAI’s need to monetize their offering and create bespoke applications for their technology that map to specific use cases and (in some cases) the job descriptions of actual humans is driving announcements like this,” AI expert John McNeil said in an email. “But here’s the truth: OpenAI is a service that its customers must derive useful, operational value from in a real world application. Let’s see if that happens. Much of what is being described is still sitting in the realm of automated tasks and not abstract or creative thinking.”
OpenAI CEO Sam Altman has been toying with the company’s product line, and he recently tweeted a proposal on social media platform X to give subscribers “credits” they can cash in for services rather than fixed subscription pricing.
“An idea for paid plans: Your $20 plus subscription converts to credits you can use across features like Deep Research, [GPT-]o1, GPT-4.5, Sora, etc.,” Altman wrote. “No fixed limits per feature and you choose what you want; if you run out of credits you can buy more. What do you think? Good/bad?”
OpenAI’s potentially wide-ranging offerings reflect a belief by many that within a year or two, we may see thousands of AI agents for every professional job category, from chip designers to high-end sales professionals. In fact, Google claims 30% of its code is now generated with AI agents, says AI agent researcher Mark Stansberry.
“The dramatic pricing jump from a $20 plan to $20,000 suggests that AI is evolving from a commoditized service into a tool capable of creating significant value — a shift that could transform traditional pricing models,” Ben James, founder of 404, said in an email. “The future of innovation will not come from AI working in isolation but from AI and humans working together to achieve more than either could alone.”
The concept of a $20,000-per-month “research scientist” with PhD-level AI could make a company’s $500,000-a-year scientist “much more productive,” Rocketlane CEO Srikrishnan Ganesan said. “They can do 2x the work they usually do,” he said in an email. “Or maybe 5x! Then the 240K/yr [expenditure on OpenAI’s high-priced AI model] makes so much sense.”
But Steven Dickens, CEO and principal analyst at HyperFRAME Research, doesn’t see any sense in the high price point. “The delta between $200 and $20,000 just doesn’t make commercial sense. Is this more a play to placate markets worried about long-term commoditization?” he wrote in an email. “What other SaaS service commands that price point??? I don’t know of any.”
In January, OpenAI introduced Operator, its first AI agent, that can automate simple web-based tasks such as grocery orders. It is available with a $200-per-month ChatGPT Pro subscription, currently OpenAI’s most expensive plan. [In October, Anthropic debuted its autonomous AI agent that can do web-based tasks for its Claude chatbot.]
As OpenAI explores higher-priced models, the company’s former head of public policy accused the company of “rewriting” its AI safety and alignment history during the launch of Chat GPT-2 in early February.
In a blog post, Miles Brundage criticized the company for rewriting the “history of GPT-2 in a concerning way.”
