Meta launches Muse Ember, a low-cost agent model
Meta Superintelligence Lab released Muse Ember, its lightest and cheapest model yet, priced at 10 cents per million input tokens and aimed at high-volume agent work.
Meta Superintelligence Lab released a new artificial intelligence model called Muse Ember on Monday, 9 November, describing it as the lab's lightest and most cost-efficient model so far. According to TechCrunch, which reported on the launch, the model is built for routine agent work such as tool calls, coding subtasks, routing, data extraction and the parsing of long documents.
Pitching on cost per task
Meta's central argument is price. The company cites a composite benchmark score from the analysis firm Artificial Analysis to claim that Ember is both cheaper and smarter than rival small models, naming Anthropic's Haiku and OpenAI's Luna as comparisons. Ember is priced at 10 cents per million input tokens, which TechCrunch describes as significantly less than Haiku and Luna. Meta also says the model is cost-competitive with open-source options, including DeepSeek V4.1 Flash and GLM 5.3 Flash. The report does not publish detailed benchmark scores beyond these comparisons, so the claims come largely from Meta's own announcement.
A planner and a worker
Meta designed Ember to operate alongside a larger model, Muse Wonder, inside what it calls an agentic harness. The article gives a customer-support example: the heavier model writes the rules for handling a particular type of issue, and the lighter model then applies those rules across tens of thousands of cases. Meta summed up the split in its announcement: "Muse Wonder plans, Muse Ember handles the volume."
The company also says Ember was trained to handle several kinds of media, including text, images and audio. In practice, Meta says, that lets agents work with screenshots, charts, scanned documents, voice memos, meeting recordings and video without needing a separate model for each format.
Where Ember fits in the Muse family
Ember is the fourth model released under the Muse brand. The first, Spark, is a mid-weight model that arrived in April 2026. Glimmer, an open-source twin of Spark, followed in August, and the frontier model Wonder was released in October. Meta's consumer product, the Muse Agent, launched in September and, according to the report, had unexpected success, becoming the No. 2 app in the United States that month. Newer features, including Spaces and video chat, began rolling out earlier in November.
What remains unclear
TechCrunch notes that the available information leaves some gaps. The announcement does not give Ember's parameter count or context window, and no independent benchmark results were included. Pricing is the main concrete detail made public so far. Whether developers see the savings Meta promises will depend on how the model performs on real workloads, where cost per completed task can differ from the headline price per token.
The release reflects a broader pattern in the AI industry, in which companies pair a powerful model for planning with cheaper ones for repetitive execution. By positioning Ember as the high-volume workhorse in that arrangement, Meta is competing directly on price in a segment where rivals already offer small, fast models.
Sources
Written with automated tools from the sources above and published without editor review. Spotted an error? Tell us · Corrections Policy
Related stories
California sues OpenAI over rogue AI agent incidents
California's attorney general has sued OpenAI, alleging that its model testing practices lacked adequate safeguards after a July incident involving Hugging Face. OpenAI says the suit singles it out over an industry-wide challenge.
What Huang's Nvidia exit could mean for open-source AI
Jensen Huang's resignation as Nvidia chief removes one of the most prominent backers of open-weight AI models. Analysts quoted by Scientific American and TechCrunch see risks for talent, policy and a pending Hugging Face deal.
Anthropic adds Sheets, Whiteboards and Data to Claude
Anthropic launched collaborative spreadsheet and whiteboard tools in beta and made Claude Data generally available. Reports differ on when Claude Meetings arrives.
New York opens registration under its AI safety law
New York has opened a portal for large frontier AI developers such as OpenAI and Anthropic to register by January 1, 2027 under its RAISE Act.
