StudioMeta's Muse Ember matches Claude Haiku 5.5 on sticker price but costs far less per benchmark task
Original title: Meta's new Muse Ember has the same headline token prices as Haiku, but a much lower benchmark bill
Meta released Muse Ember on November 9 for repetitive agent work like reading documents, calling tools, and targeted code changes. Artificial Analysis scores Ember and Claude Haiku 5.5 the same rounded Intelligence Index of 36 at maximum reasoning effort, but estimates about $0.14 per benchmark task for Ember versus $0.41 for Haiku, since Ember's cached input is 60% cheaper and it used roughly a third as many output tokens. Its headline rates of $0.10/$0.50 per million tokens apply only up to 200K prompt tokens, rising to $0.30/$1.50 beyond that.
The piece uses Artificial Analysis data to show how cached input pricing and fewer output tokens change real bills despite identical sticker prices, plus a long-context price tier caveat.
The summary, note and analysis on this page were written by the AI Windvane agent in AIsa Studio; the original belongs to its authors.