Meta launches a new AI coding model with ‘very aggressive’ pricing, CEO Mark Zuckerberg says

• On Thursday, Meta launched a new model, Muse Spark 1.1, for coding and AI agents.
• CEO Mark Zuckerberg said pricing will be “very aggressive,” calling out other labs for high costs.
• The move comes as companies throttle their employees’ AI spending due to huge bills.

Meta could spark a price war in the booming AI coding market.

The tech giant announced its latest AI model, Muse Spark 1.1, on Thursday, saying it performs well on industry tests for coding and AI agents. It’s Meta’s first AI model that it charges users for.

In comments on X, Meta CEO Mark Zuckerberg said the model has a “very low price.”

Zuckerberg called out other AI companies for pricing their chatbots at “very extreme” levels in comments to Bloomberg. He told Bloomberg that the model outperformed Google’s Gemini in several categories, including agents, coding, and other capabilities.

“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,” Zuckerberg told the outlet.

The model, which isn’t fully available to developers yet, marks the latest milestone for Meta’s AI efforts — and shows the company intends to compete on price.

If Meta’s new AI models can compete with widely-used coding tools from rivals like Anthropic, OpenAI, and Cursor, that could represent a huge new source of revenue. Meta’s stock was up nearly 2% on Thursday.

Meta’s chief AI officer, Alexandr Wang, told CNBC that the company will charge $1.25 per million input tokens and $4.25 per million output tokens — lower than the pricing for Google and OpenAI’s flagship models.

The cost of using AI has become a growing concern for companies as employees incorporate the technology into more of their day-to-day work. Companies have been throttling their employees’ use of AI in recent months as vibe coding takes off. Coinbase, for example, now limits its engineers’ weekly AI spending to $500 to $5,000 a week.

Meta quoted one of its customers, AI coding startup Cline, saying that the new AI model’s price point makes it easy to run heavy AI coding tasks at scale.

“That combination is rare, and it’s exactly why we wanted Cline developers to have access early,” Saoud Rizwan, the Cline CEO, said on Meta’s website.

Meta is spending massive amounts of cash on AI, raising its capital expenditure guidance for this year to $125-$145 billion, up from a previous estimate of $115-$135 billion. Meta remains highly dependent on its ads business, which accounts for about 98% of its total revenue, according to its first-quarter earnings results.

“We believe Meta is well positioned to generate ample revenue to support its spending, driven by monetization of its own AI initiatives, advertising share gains, incremental subscription revenue, an optionality of cloud offering, and fees for external use of its AI models,” BNP Paribas Equity Research senior analyst Nick Jones wrote in a note to investors on Thursday.

Joe Tigay, Equity Armor Investments portfolio manager, said that undercutting prices signals that Meta’s AI lags behind Anthropic’s and OpenAI’s.

“They are forced to compete on price and raw computing real estate rather than model superiority,” he told Business Insider.

Meta is also working on a coming AI model codenamed “Watermelon,” which Wang said has caught up to one of the latest versions of OpenAI’s ChatGPT. The model uses “an order of magnitude” more computing power than Meta’s previous model, Wang told staff last week, Business Insider reported earlier.

Meta didn’t respond to a request for comment.

Correction: July 9, 2026 — An earlier version of this story misspelled an analyst’s name. It should be Nick Jones. This story has also been updated to include Wang’s comments to CNBC on the model’s pricing.

Have a tip? Contact Charles via email at crollet@businessinsider.com or on Signal and WhatsApp at 628-282-2811. Use a personal email address, a nonwork WiFi network, and a nonwork device; here’s our guide to sharing information securely.