The new Meta AI coding agent, Muse Spark, offers low-cost software development to directly rival OpenAI and Anthropic.
Meta is rolling out Muse Code, its first dedicated coding agent, as it steps up efforts to challenge leading AI labs Anthropic and OpenAI.
Muse Code is the latest flagship release under AI Chief Alexandr Wang, who leads Meta Superintelligence Labs and oversees foundation model development. Wang joined Meta last June as a central figure in CEO Mark Zuckerberg’s strategy to overhaul the company’s artificial intelligence division.
Similar to Anthropic’s Claude and OpenAI’s Codex, the new tool simplifies app development within a single user interface, enabling users to manage fleets of AI agents across the software pipeline.
“Muse Code is a terminal coding agent, like many others on the market,” Wang explained. “You can install it with a single command and use it to execute complete software engineering tasks from planning changes and writing code to validating results.”
Available in a public beta preview, Meta’s new coding agent works alongside its latest AI model, Muse Spark 1.2. Like its July predecessor (MuseSpark 1.1), this updated model is heavily optimized for programming tasks.
What sets Muse Spark 1.2 apart is that it was co-developed and trained directly alongside Muse Code a joint training process that Wang says significantly boosts overall coding performance.
Developers can access Muse Code via a pay-as-you-go plan featuring the same API pricing as the Muse Spark 1.1 release $1.25 per million input tokens and $4.25 per million output tokens.
See also: Google AI Reshuffle: Demis Hassabis Steps Down as DeepMind CEO as Jeff Dean Departs
Wang noted that Meta is also offering a contributor tier at a significantly lower cost, calling it “more than 10 times cheaper” than the pay-as-you-go option.
“As part of that tier, users opt in to help improve the model using industry-standard practices aligned with other coding agents,” Wang explained.
