Meta takes on OpenAI and Anthropic with Muse Ember AI model: Features, benchmarks and pricing
LiveMint

Meta launched Muse Ember, a small, cost-efficient AI model designed to handle repetitive tasks and agentic workflows for businesses.
Meta has announced Muse Ember, its smallest AI model yet, designed specifically to handle large-scale, repetitive tasks and agentic workflows. Aimed at businesses and developers, the model is built to automate routing, coding subtasks, document processing, and data extraction at a lower cost than competitors.
According to Meta CEO Mark Zuckerberg, Ember delivers greater intelligence per dollar than rival models such as Anthropic’s Claude Haiku 5.5 and OpenAI’s GPT-6 Luna. The model natively supports text, images, audio, and video, boasting a 1 million token context window and up to 512,000 output tokens. It is priced at $0.10 per million input tokens and $0.50 per million output tokens.
Benchmark results released by Meta show Ember scoring 40.9% on Terminal-Bench 4.0, outperforming Claude Haiku 5.5 and GPT-6 Luna, though Claude Haiku 5.5 scored higher on FrontierCode 2.0-AA. Ember works alongside larger models in the Muse family, with Meta explaining that larger models handle complex planning while Ember handles execution.
Following its launch, Muse Ember, along with other Muse models, has been made available through Google Cloud and Oracle Cloud. Meta plans to continue developing efficient AI models to enable advanced capabilities at scale.
Why it matters: Muse Ember provides businesses with a cheaper, highly efficient option for executing routine AI workloads and agentic workflows at scale.




