Meta offers substantial discounts for users sharing data with its AI model

Meta is offering significant cost reductions for users who share their data with its new Muse Spark AI model. Under the contributor pricing model, input tokens cost 10 cents per million instead of $1.25, and output tokens cost 20 cents per million instead of $4.25. Mario Zechner, developer of the Pi harness, noted that data collection was key to improving coding agent capabilities. Arvind Narayanan, a Princeton professor, highlighted that large companies often avoid sharing data for model training. Meta’s pricing guide suggests the contributor tier aims to lower barriers for prototyping and testing. Anthropic and OpenAI have also introduced price cuts for their models, indicating growing competition in the AI sector. The initiative to track employee computer usage, launched earlier this year, faced internal criticism and was paused in June. Meta did not comment on the new pricing model. Narayanan suggested the contributor model could encourage companies to be more selective about data sharing. The move aligns with broader industry trends toward price competition and data-driven model improvements.