Meta has fired the opening shot in an AI price war. The company launched Muse Spark 1.1 and the Meta Model API, its first paid developer tier, with pricing designed to undercut rivals.
The new model costs $1.25 per million input tokens and $4.25 per million output tokens. That's roughly 25% cheaper than comparable offerings from OpenAI and Anthropic. New accounts receive $20 in free credits.
Muse Spark 1.1 is a multimodal model built for complex, multi-step tasks. It can orchestrate AI agents and integrate with platforms like GitHub. It supports a context window of one million tokens.
This marks Meta's shift from its open-source Llama releases toward monetizing its AI infrastructure. The API is compatible with OpenAI's SDK, allowing developers to switch without rewriting their code.
CEO Mark Zuckerberg called the pricing "very aggressive and attractive." The initial preview is available only to US developers. Market reaction was positive, reflecting investor confidence in Meta's broader data center expansion plans.
For blockchain developers, the lower costs could make running complex analytical workflows more affordable. Features like multi-agent orchestration are relevant for automated portfolio management and smart contract review.
However, building on Meta's centralized API introduces platform dependency, which contrasts with the decentralization principles central to Web3.