META LAUNCHES MUSE CODE CODING AGENT IN EARLY BETA

Meta has announced an early beta release of Muse Code, a terminal-based coding agent designed to assist software engineers with complex tasks across large codebases. The tool is powered by Muse Spark 1.2, a new version of Meta's AI model offering what the company describes as improvements in code generation, complex debugging, codebase understanding and end-to-end developer workflows. Muse Code can write code, plan changes and validate results based on text prompts, and can manage multiple sub-agents to delegate tasks more efficiently.

The tool competes directly with Anthropic's Claude Code and OpenAI's Codex. Meta has chosen pricing as a differentiation strategy, offering standard pay-as-you-go rates of $1.25 per million input tokens and $4.25 per million output tokens. The company also plans to offer a contributor tier at substantially lower rates—$0.10 per million input tokens and $0.20 per million output tokens—for users willing to provide feedback to improve the agent. These prices undercut Anthropic's Sonnet 5 model, which costs $3 per million input tokens and $15 per million output tokens.

Meta demonstrated Muse Code through sample projects including an interactive photon sphere model and a Plants vs. Zombies adaptation, alongside a demonstration video showing the tool building a webpage from an MP4 file. The pricing structure reflects Meta's strategy to compete on cost as it expands its presence in the AI tools market. Alexander Wang, Meta's Chief AI Officer, confirmed the contributor tier pricing in an interview with CNBC.