.png){kind=logo}
Meta Superintelligence Labs has launched Muse Ember, a new lightweight artificial intelligence model designed to handle repetitive tasks performed by AI agents. The company describes it as its lightest and most cost-efficient model yet, focusing on tool calls, coding subtasks, routing, data extraction, and long-document processing.
Announced on November 9, Muse Ember is designed to deliver accuracy at scale while keeping operating costs low. Meta is positioning the model as an execution-focused system that can work alongside larger AI models rather than replace them.
Meta has priced Muse Ember at USD 0.10 per million input tokens, making it significantly cheaper than frontier models such as Anthropic's Haiku and OpenAI's Luna. The company also highlighted its cost competitiveness against open-source alternatives, including DeepSeek V4.1 Flash and GLM 5.3 Flash.
To demonstrate its pricing and performance, Meta pointed to Artificial Analysis's composite benchmark score, arguing that Muse Ember offers a stronger balance between capability and cost.
The model is intended for workloads involving thousands of repeated operations. For instance, in a customer-support system, a larger AI model could establish rules for handling a particular complaint. At the same time, Muse Ember applies those instructions across tens of thousands of customer cases.
Meta summed up this approach by saying, "Muse Wonder plans, Muse Ember handles the volume."
Also Read: Meta’s Muse AI Turns Ideas Into Custom In-App Tools with Spaces
Beyond its execution capabilities, Muse Ember has been trained to process different types of media, including text, images, and audio.
According to Meta, AI agents using the model can analyze screenshots, charts, and scanned documents. They can also work with voice memos, meeting recordings and videos without requiring a separate model for each format.
This multimodal capability could make Muse Ember useful for businesses deploying AI agents across different workflows, particularly those involving large volumes of information and repetitive tasks.
Muse Ember is the fourth model introduced under Meta's Muse brand. The lineup includes Spark, a mid-weight model released in April, followed by Glimmer, its open-source counterpart, in August.
Meta launched its frontier model, Wonder, in October. The company's flagship consumer AI product, Muse Agent, arrived in September and saw unexpected success. It introduced features such as Spaces and video chat earlier this month.
With Muse Ember, Meta is expanding its AI portfolio towards businesses and developers looking to run agent-based operations at scale without incurring the higher costs associated with larger models.