Meta has introduced a novel pricing model for its latest AI model, Muse Spark, which incentivizes users to share their prompts and model outputs to aid in the development of future iterations. Under this “contributor pricing” scheme, users receive substantial discounts—approximately 95% off the standard rates—for contributing their usage data. Specifically, the cost for one million input tokens drops from $1.25 to 10 cents, while output tokens decrease from $4.25 to 20 cents per million under the contributor model.
This initiative comes in the wake of Meta’s previous struggle to gather training data, including an internally criticized employee monitoring program that was halted in June. The company has not commented on the new pricing structure. The move reflects a broader industry trend where model builders, particularly those focusing on agentic tools, seek user data to enhance tool performance, despite challenges in accessing comprehensive digital traces of professional workflows. Princeton computer science professor Arvind Narayanan suggests that Meta’s compensation framework may encourage companies to more carefully distinguish between proprietary and shareable data, potentially influencing competitive dynamics as other AI leaders like Anthropic and OpenAI also adjust their pricing strategies.