Meta has launched a contributor pricing model for its Muse Spark model — designed for coding and other agent operations — that sharply reduces token costs for users who allow the company to use their interactions for model development. Under the standard agreement, 1 million input tokens cost $1.25, while the contributor tier drops that to 10 cents. For output tokens, the standard $4.25 per million falls to 20 cents.

The move comes as Meta has faced difficulty obtaining training data. An internal initiative to track employee computer usage, launched earlier this year, drew widespread criticism and was paused in June. Meta did not respond to a question from TechCrunch about the new pricing model.

User interaction data is considered vital for improving agentic AI tools. Mario Zechner, developer of the open-source harness Pi, told TechCrunch that a significant jump in coding agent capabilities between April and October 2025 was driven by Claude Code storing user sessions for reinforcement learning. However, evaluating and improving tools beyond software engineering remains challenging, as many professional workflows lack digital traces.

Princeton computer science professor Arvind Narayanan noted that large companies tend to avoid sharing data for training, opting for enterprise plans with higher costs but stricter data controls. Meta's pricing guide frames the contributor tier as a way to "lower the barrier to entry for prototyping, testing integrations, and scaling experiments where training on your data is acceptable." Narayanan suggested this could encourage companies to more carefully distinguish between proprietary and sharable data.

The pricing strategy also coincides with broader price competition among frontier AI labs. Anthropic's new Fable and Mythos models, released the day before, reduced costs for cached tokens, while OpenAI implemented major price cuts at the end of July.