Meta Platforms Inc. unveiled a new API for its AI models on Thursday, with CEO Mark Zuckerberg promising pricing that is “aggressive and attractive” — about 25% of the cost of comparable offerings from OpenAI and Anthropic. The move signals an intent to undercut dominant players in the rapidly expanding market for generative AI tools.
The API supports Muse Spark 1.1, the latest iteration of Meta's in-house model series first introduced in April. The company describes the update as a “step-change” from the first generation, claiming it can handle advanced coding tasks such as detecting and fixing complex bugs. It also supports end-to-end agentic workflows, including multi-agent systems, across applications.
Muse Spark 1.1 boasts native multimodal perception for images, videos, and documents, broadening its utility beyond text. The model was refined using developer feedback, reflecting Meta's push to compete in a space dominated by OpenAI and Anthropic. Zuckerberg's pricing strategy aims to lure cost-sensitive developers away from rivals.
By slashing API costs, Meta could accelerate adoption of its AI ecosystem, challenging the market share of established players. The move may pressure competitors to adjust pricing, potentially reshaping the economics of enterprise AI tools. Developers evaluating model performance versus price will play a key role in determining whether Meta's gambit succeeds.
Critics note that cost alone may not sway enterprises; model reliability, safety, and ecosystem lock-in remain significant factors. Meta's track record on data privacy could also deter some adopters, particularly in regulated industries.