Zuckerberg plans 'aggressive' pricing strategy for Meta paid AI model
In the fiercely competitive AI tool market, Mark Zuckerberg hopes to win with price advantage. Meta Platforms Inc. Released its most advanced artificial intelligence model, Muse Spark 1.1, and launched a paid version for developers for the first time. This is the first time Meta has charged an enterprise a model usage fee, and it has also opened up a new source of revenue for the company. Zuckerberg said in an interview ahead of launch that the model would be one of the lowest-priced options on the market. "Because this is not an open source model, I think this is the first time we're really serious about launching an API," Zuckerberg said. "The pricing will be very competitive and very attractive." Zuckerberg said that the biggest improvement of the new model is the ability of the agent. Agents are one of the hottest directions in the field of AI this year. They usually refer to systems that can complete multi-step tasks on behalf of users. He said that Muse Spark 1.1 has reached "industry leading or very close to leading" levels in terms of agent reasoning and tool usage. He also said that the model has also significantly improved in terms of programming capabilities, and Meta employees have used it internally to develop products and features for various applications. Meta will also launch a new Meta Model API system for charging developers. Zuckerberg said its API price is about 25% of the top models from OpenAI and Anthropic. Developers can use the Meta model for free, but after reaching a certain token usage threshold, they will need to pay. Zuckerberg said: "Some other AI labs have very high pricing and high profit margins." He emphasized that Meta's strategy is to let as many people as possible use its AI technology. "We believe that we have the ability to provide cutting-edge, or high-level, artificial intelligence at a lower cost." Despite the huge investment, Meta models have historically performed worse in tests than Anthropic, OpenAI and . However, Zuckerberg said that the competitiveness of Muse Spark 1.1 has been significantly enhanced and is better than the Google Gemini model in tests in multiple categories such as agents, programming and multi-modality. He said: "This is a quite interesting milestone. I think this may be the first time in my memory that the Meta model comprehensively surpasses all Google models." Zuckerberg’s investment in the AI race has triggered a series of major strategic and resource adjustments over the past year. After a disappointing model launch in the spring of 2025, he personally stepped in to rebuild Meta AI Labs, including hiring former Scale AI CEO Alexandr Wang to lead a new team, followed by massive layoffs and multiple rounds of internal reorganization. Meta, once firmly committed to open source AI models that are free to external developers, has now shifted to prioritizing the development of chargeable closed-source models, such as Muse Spark 1.1. This shift almost meant that the technology needed to be rebuilt from scratch. Zuckerberg said he was satisfied with the current progress of the AI lab. "Overall, our performance is better than expected." He said that Meta is still lagging behind large AI labs such as Anthropic and OpenAI, but the company also has a new model code-named Watermelon under development, which he believes will help Meta "further advance the frontier of intelligence." He declined to disclose the release date of Watermelon, saying only that the current focus is on improving quality. Zuckerberg does not agree that AI will eventually become a homogeneous commodity, that is, the functions of various models will converge and be difficult to distinguish from each other. He cited Anthropic's latest model, Mythos, as an example, saying that the model has been restricted due to US national security concerns, which shows that companies have begun to control some technologies rather than widely opening them up. He said: "These capabilities have not actually proliferated, nor are they widely open to everyone. Anthropic basically kept the strongest model to itself and only released a relatively simplified version. Of course, they may have various reasons for doing this. But at least in my opinion, this does not mean that the world is, or will inevitably, move in the direction of widespread openness of AI capabilities."