Meta Launches Muse Ember AI Model for Coding and Automated Tasks

Meta has introduced Muse Ember, a new AI model aimed at developers and businesses that use AI agents to automate everyday tasks. The model is designed to handle smaller jobs within larger workflows, such as reading files, running code tests, extracting information and calling external tools. Meta says Muse Ember can deliver performance comparable to larger models at a lower cost, making it suitable for tasks that agents need to perform repeatedly. It is now available through the company's developer platforms.

Meta Muse Ember Features, Benchmarks

Muse Ember is designed for the smaller tasks involved in running AI agents, Meta said in a blog post. These include reading documents, making changes to code, running tests and retrieving information. It can also use external tools and handle routine business work, such as finding details in company records and updating them across applications, the company added.

The model supports a 1-million-token context window and can generate up to 512,000 output tokens. It accepts text, images, audio and video as inputs, allowing it to work with different types of content. Developers can use it to analyse lengthy documents, review screenshots and work with large codebases without having to split them into several smaller requests.

Meta has shared benchmark results comparing Muse Ember with Anthropic's Claude Haiku 5.5 and OpenAI's GPT-6 Luna. The model scored 36 points in the Artificial Analysis Intelligence Index, matching Claude Haiku 5.5 and scoring higher than GPT-6 Luna, which received 28 points. Meta also claims that Muse Ember delivers performance similar to Claude Haiku 5.5 at around a third of the cost.

In AutomationBench-AA, which tests how well AI models handle automated tasks, Muse Ember scored 37.8 percent. This was higher than GPT-6 Luna's 20.2 percent and Claude Haiku 5.5's 13.9 percent. It also scored 40.9 percent in Terminal-Bench 4.0, slightly ahead of Claude Haiku 5.5 at 39.2 percent. Claude's model, however, performed better in FrontierCode 2.0-AA, scoring 34.9 percent against Muse Ember's 27.2 percent.

Muse Ember is the first model Meta says it has trained using Muse Wonder Full. The company also put it through safety evaluations covering cybersecurity threats, harmful instructions and other potential risks. Meta said the model did not cross any of its high-risk thresholds.

Meta Muse Ember Availability

Meta has released Muse Ember through its Model API and Muse Code, where developers can use it muse-ember-1.0. The company charges $0.10 (roughly Rs. 9.36) per million input tokens and $0.50 (roughly Rs. 47) per million output tokens. It charges $0.004 (roughly Rs. 0.37) per million cached input tokens. Meta has also added support for the Responses API and made the model available on Google Cloud and Oracle Cloud.