Key Highlights

  • Meta has launched Muse Code, its first coding agent, aiming to compete with Anthropic's Claude and OpenAI's Codex in the AI-assisted software development market.
  • The tool is led by AI chief Alexandr Wang, who heads Meta Superintelligence Labs, and works alongside Meta's newest model, Muse Spark 1.2.
  • Muse Code is priced at 1.25 dollars per million input tokens and 4.25 dollars per million output tokens under its pay-as-you-go option, matching Muse Spark 1.1 pricing.
  • A contributor tier offers access at more than ten times cheaper than the pay-as-you-go tier in exchange for opting in to help improve the model, alongside a newly introduced zero-data-retention option for enterprise users.

Meta Enters Coding Agent Market With Muse Code

Meta (NASDAQ:META) has begun rolling out Muse Code, its first coding agent, as the company works to establish a stronger position against Anthropic and OpenAI in the software development tools market. The release is the latest from AI chief Alexandr Wang, who leads Meta Superintelligence Labs and oversees the company's foundation model development. Wang joined Meta last year as part of CEO Mark Zuckerberg's broader effort to rebuild the company's artificial intelligence strategy. Muse Code functions as a terminal-based coding agent, allowing developers to install it with a single command and use it to handle software engineering tasks including planning changes, writing code, and validating results, similar in structure to existing tools from competing AI labs.

Tool Built Alongside Meta's Newest Coding-Focused Model

Muse Code is currently available in a beta, or preview, version and operates alongside Muse Spark 1.2, Meta's latest model, which specializes in coding tasks. Unlike the prior Muse Spark 1.1 release from July, the newest model version was developed and trained in tandem with Muse Code itself, a process Wang said improves overall coding performance. Developers can access Muse Code through a pay-as-you-go pricing structure that mirrors Muse Spark 1.1's API pricing, at 1.25 dollars per million input tokens and 4.25 dollars per million output tokens. A separate contributor tier offers substantially lower costs, more than ten times cheaper than the pay-as-you-go option, in exchange for developers opting in to help improve the model through data sharing.

Pricing Strategy Positions Meta as Low-Cost Alternative

Wang said Meta is differentiating Muse Code and the broader Muse Spark model family primarily through pricing rather than through claims of leading-edge performance relative to Anthropic's Claude Code or OpenAI's Codex. The company has also begun accepting requests for zero-data retention, an option under which Meta would not retain developer data for model improvement or other purposes, which Wang described as an important enterprise feature. This combination of competitive pricing and enterprise-focused privacy options may help Meta attract cost-sensitive developers, though the company's ability to gain sustained share is likely to depend on how its coding performance compares as rivals continue releasing updated models of their own.

Conclusion

Muse Code marks Meta's formal entry into the AI coding agent market, extending the company's broader effort under Alexandr Wang to close the gap with Anthropic and OpenAI in agentic AI tools. By pairing the new tool with its latest Muse Spark 1.2 model and emphasizing lower pricing alongside new enterprise privacy options, Meta is positioning itself as a cost-competitive alternative rather than claiming outright technical superiority. The tool's reception among developers, still in beta, may offer an early signal of whether Meta's pricing-led strategy can meaningfully shift market share in a competitive field.