Hugging Face makes six-hour ML runs a chat task

Hugging Face launched ML Intern in its open-source repository, an agent that turns a chat request into model search, training, monitoring, reports, and demos; one launch demo ran for about six hours for less than $0.50. In The Decoder’s account of the release, the budget is approved before compute starts, but the cost example remains vendor-reported. Builders following the autoresearch loop’s case for AI-run experiments should measure actual compute and review costs before trusting that ceiling.