[Meta Launches Muse Ember: 17 Times Cheaper Than Its Own Spark, Matching Claude Haiku 5.5] Beating AI News: Meta has released Muse Ember, its lightest AI model, specifically designed for tasks such as repetitive tool calls, code modifications, and document analysis. Ember supports text, image, audio, and video input, and its context window holds up to 1 million tokens. On Artificial Analysis's Intelligence Index, Muse Ember scored 36 points, the same as Claude Haiku 5.5, but its cost per task is only $0.14, about one-third of Haiku's. Compared to Meta's own Muse Spark 1.3, Ember is only 1 point lower, but its cost is drastically reduced to about 1/17th. The key to its low cost lies in its efficient token utilization. Ember and Haiku 5.5 have the same standard input and output unit prices, but Ember's cache read price is 60% lower, and the output tokens generated when completing evaluation tasks are only about one-third of Haiku's. Meta plans to have the large model Muse Wonder handle complex tasks, then delegate a large amount of repetitive work to Ember. For example, Wonder will develop code migration plans, with multiple Ember agents modifying files and running tests in parallel. However, Ember also has significant weaknesses. In the Artificial Analysis knowledge test, it only answered 21.1% of the questions correctly, rejecting 64% of them. In the more complex engineering programming test FrontierCode 2.0, its score was 27.2%, lagging behind Haiku 5.5's 34.9%. Currently, Ember is available on Muse Code and the Meta Model API. The standard price is $0.10 per million input tokens and $0.50 per million output tokens. However, the price triples once the input in a single request exceeds 200,000 tokens. [Original Link]