Meta is cutting Muse Spark prices by roughly 95% for customers who let it train on their prompts and the model’s outputs. The contributor tier runs 10 cents per million input tokens and 20 cents per million output, against $1.25 and $4.25 on a standard agreement. Meta has had trouble sourcing training data by other means, pausing an internal effort to track employees’ computer use in June after it drew wide criticism inside the company, and it didn’t answer TechCrunch’s questions about the tier. Princeton computer scientist Arvind Narayanan wrote on X that big companies stay on metered enterprise plans even when consumer subscriptions cost 10 to 20 times less, mostly over data retention and IT governance.
