OpenAI, Meta, SpaceXAI compete for more cost-efficient AI models

[NEW YORK] Three prominent artificial intelligence developers released new models over the past week. They all promise to be more advanced, but their biggest immediate selling point may not be what they can do but how little they charge to do it.

OpenAI said its most advanced offering, GPT-5.6, is designed to complete more work while using significantly fewer tokens, a unit of data processed by AI models. This will make the software far more cost efficient for customers. 

Grok 4.5, from Elon Musk’s SpaceXAI, is billed as having twice the token efficiency as comparable models from other firms.

And Meta Platforms is making the pricing for its Muse Spark 1.1 very “attractive”, chief executive officer Mark Zuckerberg told Bloomberg.

The renewed emphasis on cost coincides with business customers scrutinising AI spending.

Earlier in 2026, firms encouraged employees to outdo one another by using AI as much as possible, a practice known as tokenmaxxing.

But in recent months, some companies have imposed tighter limits after being hit by sticker shock, in part due to developers like Anthropic switching to usage-based pricing rather than simply charging a flat subscription fee.  

Gautier Cloix, CEO of Paris-based AI startup H Company, said he has spoken with a number of executives whose businesses have racked up significant bills after using models from OpenAI and Anthropic.

One CEO showed him an invoice indicating a month of AI model usage cost millions of dollars, Cloix said.

SEE ALSO

Cheaper AI is better: Soaring bills are reshaping how businesses choose models

As ‘tokenmaxxing’ and AI costs take off, companies turn to private token factories: Dell

The AI bubble will surely pop sometime, but it’s an irrational exercise to predict exactly when

‘Tokenmaxxing’ your way to a zero-day work week

More

“Companies are spending a lot more than they used to,” said Gil Luria, head of technology research at DA Davidson & Co. “As they see these costs get out of control, they’re starting to ask questions about efficiency.” 

As a result, top AI developers must now find ways to maximise value for current and potential customers, without undercutting themselves to the point where they would struggle to recoup the hundreds of billions of dollars they have invested in chips and data centres.

Meta, which benefits from having a lucrative online advertising business, is prepared to be “aggressive”, according to Zuckerberg.

“The pricing from some of the other labs is very extreme and has very high margins,” he said in the interview. “We think that there’s a real ability to be able to offer frontier or very high-level intelligence at a much more affordable cost.”

OpenAI may have less wiggle room but also sees the need to be competitive on costs. “Every enterprise now is thinking about spend and the value they’re getting in exchange for AI, and this is what we really want to do,” OpenAI CEO Sam Altman said in an interview with CNBC on Thursday (Jul 9).

The rhetoric is notably different from a year or so ago, when OpenAI executives were publicly musing about one day charging thousands of dollars for monthly subscriptions to top-tier AI models to better reflect the growing value they provide to businesses. 

Now, in addition to introducing more efficient new models, OpenAI has also been taking steps to help companies manage their AI outlays. Last month, the ChatGPT maker introduced credit usage analytics and updated spending controls. 

As businesses get more cost-conscious, they are “looking for other solutions”, Luria said. And there is no shortage of alternatives.

Chinese tech companies like DeepSeek have flooded the market with more affordable open AI models. While these services still lag behind the most cutting-edge options from US firms, the software is good enough to field many day-to-day duties.

Some users are turning to model routing services, which allow them to seamlessly select from hundreds of AI models for various tasks to ensure better prices. One such service, OpenRouter, raised more than US$100 million in funding in May to meet demand. 

By emphasising cost efficiency, AI developers may also be able to put more pressure on Anthropic, viewed by many as the frontrunner of the moment.

Anthropic’s Opus and Fable models rank among the most expensive on a cost-per-task basis, according to data from Artificial Analysis, a benchmarking service. 

Musk specifically took aim at Anthropic in a post touting Grok 4.5 this week. “It is an Opus-class model,” he wrote on social media, “but faster, more token-efficient and lower cost.” BLOOMBERG