Meta has made a strategic and significant move into the artificial intelligence market by making a proprietary AI model available to businesses for the first time. In a notable release on Thursday, July 9, 2026, the company introduced Muse Spark 1.1, promising up to 75% lower usage costs compared to cutting-edge AI models offered by rivals like OpenAI and Anthropic. The initiative reflects Mark Zuckerberg’s vision of democratizing access to AI by encouraging developers to build their solutions on Meta’s platform.
The reason behind aggressive price reduction
This launch marks an important shift in Meta’s approach, which has previously been a fervent supporter of open source AI. With Muse Spark 1.1, the company enters the proprietary API business sector, seeking to generate revenue. Mark Zuckerberg highlighted that Meta is only charging about a quarter of what OpenAI and Anthropic demand for their services. The intention is clear: to make artificial intelligence so accessible that a vast ecosystem of developers opts for Meta’s platform.
Muse Spark 1.1 cost comparison with rival models
Meta’s pricing strategy is one of the most striking points of the launch, with the aim of undermining the competition and attracting developers. The promised savings are substantial and could redefine the cost landscape for building AI applications.
Learn more: Discover the main signs that your WhatsApp account is being used without permission
- Entry Tokens:While competitors charge between $5 and $10 per million input tokens, Meta offers its services at a significantly lower cost.
- Output Tokens:In terms of exit tokens, prices charged by rivals range from US$30 to US$50 per million tokens, a range that Meta seeks to drastically reduce.
- Performance versus cost:Independent research has indicated that Muse Spark 1.1 can operate at approximately one-tenth the cost of GPT-5.5 and have input costs 75% lower than Anthropic’s Claude Opus 4.8, as well as having 83% lower output costs.
Meta’s ambitious AI infrastructure investment plan
Meta’s offensive in the field of AI is not just about competitive pricing; it is supported by massive investment in infrastructure. The company has allocated between US$125 billion and US$145 billion in spending capital for 2026, the largest amount in its history. This billion-dollar investment aims to build and strengthen its AI computing capacity.
A crucial part of this plan involves the development of a custom artificial intelligence chip, dubbed “Iris,” in collaboration with Broadcom and manufactured by TSMC. The goal is to reduce Meta’s dependence on external vendors like Nvidia and significantly lower AI inference costs. An internal memo revealed plans to start Iris production in September and double the company’s computing capacity to 14 gigawatts, reinforcing Meta’s ambition to lead the AI race.
How Meta’s new pricing could impact the artificial intelligence market
Meta’s decision to enter the proprietary API market with such an aggressive cost structure could have broad repercussions for the artificial intelligence industry. By offering substantially lower prices, the company is not only trying to attract a massive developer base to its platform, but also putting pressure on competitors to re-evaluate their own pricing strategies.
On the same topic: Learn how to identify intrusions on WhatsApp and block scammers quickly
This movement could catalyze a “price war” in the AI model segment, making the technology more accessible to a wider range of companies and startups. In the long term, this could accelerate the innovation and adoption of AI in several industries, as the cost barrier to entry would be significantly reduced. Meta’s strategy, therefore, is not just about its own growth, but about the possible reconfiguration of the global artificial intelligence economy.
Uncertainties and the future “Watermelon” model
Despite the optimism surrounding the launch of Muse Spark 1.1, Meta faces internal challenges. On July 2, 2026, Zuckerberg admitted that the growth of AI has not accelerated according to internal expectations. Although the Muse Spark 1.1 is competitive in price and capacity, it still lags behind the GPT-5.5 in raw performance.
Meta’s real bet lies in its cutting-edge model, known internally as “Watermelon”, which is still under development. The success or failure of Watermelon will be crucial to validating the US$145 billion investment in infrastructure. If the future model delivers the expected performance, the Muse Spark 1.1’s aggressive pricing strategy will have been a smart move to build a user base. Otherwise, Meta may have started a price war that it cannot win, with significant financial implications for the company.