Meta Launches Muse Code to Challenge Anthropic, OpenAI
The coding agent pairs with Muse Spark 1.2 as Meta seeks to compete on price in a market led by Claude Code and Codex.
Topics
News
- Temasek, Seraphim Back Pixxel in $100 Million Round
- UN Rights Chief Warns Advanced AI Could Threaten Humanity
- Tata-owned JLR to Cut About 4,000 Bobs Amid Tariff, China Pressure
- Anthropic Walks Away From $6 Billion Decart Deal
- Hyundai India Targets 20% Women in Executive Workforce by 2030
- Security Breach Drains $320 Million in Bitcoin From Liquid Network
[Image source: Diksha Mishra/MITSMR India]
Meta Platforms has launched Muse Code, its first artificial intelligence coding agent, as the Facebook and Instagram owner expands into developer tools and challenges established products from Anthropic and OpenAI.
Released in beta for macOS and Linux, Muse Code is a terminal-based agent powered by Meta’s new Muse Spark 1.2 model. It can work across large software repositories to plan changes, write code and validate the results, while coordinating multiple agents on different parts of a project. Meta announcement
“You can install it with one command and then use it to take on complete software engineering tasks,” Meta AI chief Alexandr Wang told CNBC.
Wang, who leads Meta Superintelligence Labs and oversees the company’s foundation-model development, joined Meta last year as part of Chief Executive Mark Zuckerberg’s effort to rebuild its AI operations.
Muse Code uses persistent background agents that remain active throughout a session, reducing the need to repeatedly examine the same code or gather the same information. It also records model calls, tool use, approvals and edits in a local activity log, allowing work to resume from the same point after a crash.
Meta is offering standard access on a pay-as-you-go basis at $1.25 per million input tokens and $4.25 per million output tokens. A discounted contributor tier costs $0.10 per million input tokens and $0.20 per million output tokens, but requires developers to allow Meta to use their data to train and improve its models. Meta pricing
The contributor option is more than 10 times cheaper than the standard tier, making pricing a central part of Meta’s attempt to attract developers from competing services.
Wang told CNBC that Meta was seeking to distinguish Muse Code primarily through cost rather than claiming that it was more capable than Anthropic’s Claude Code or OpenAI’s Codex.
Meta is also beginning to accept requests for zero-data retention, which would allow some enterprise customers to use the service without their developer data being retained for model improvement.
Muse Spark 1.2 was trained alongside Muse Code, which Meta said helped improve its coding capabilities. The model will also be offered through OpenRouter, a platform that provides access to models from multiple AI developers.
Coding has become one of the most competitive markets for commercial AI agents as developers increasingly use them to generate software, find bugs and complete multistep engineering tasks. Anthropic has expanded Claude Code rapidly, while OpenAI offers Codex through ChatGPT and its developer products.
Muse Code also gives Meta another potential route to monetize the infrastructure behind its AI models as investors scrutinize the cost of Zuckerberg’s expansion. Meta recently raised the lower end of its 2026 capital-spending forecast to $130 billion from $125 billion, while retaining the upper end at $145 billion. Meta second-quarter results.
