Meta has introduced Muse Code, a cost-efficient AI coding tool in beta, aiming to disrupt the heavily contested agentic coding segment dominated by OpenAI and Anthropic, with a focus on large-scale codebase management and lower price points.
Meta Platforms has entered the AI coding market with Muse Code, a terminal-based agent designed to help developers plan changes, write software and verify results across large codebases. The product, launched on 5 August and now in beta for macOS and Linux, marks Meta’s clearest move yet into agentic coding tools, a segment already crowded by Anthropic and OpenAI. According to reporting by MacRumors and TechCrunch, the system uses Meta’s Muse Spark 1.2 model and can coordinate multiple persistent sub-agents to split up complex work.
Price is the sharpest part of Meta’s pitch. The standard plan is set at $1.25 per million input tokens and $4.25 per million output tokens, while users who allow Meta to use prompts and completions for training can access a far cheaper Contributor tier at $0.10 per million input tokens and $0.20 per million output tokens, according to MacRumors. That discounted pricing undercuts many rival offerings and is meant to appeal to developers who care more about cost efficiency than premium brand positioning. Alexandr Wang, Meta’s AI chief, told TechCrunch the tool can be “an incredibly good option for a lot of workflows, especially from a cost perspective.”
Meta is also leaning on performance claims. The company says Muse Code can handle large repositories by generating plans, producing code and validating outcomes, while keeping an append-only local log so it can resume after a crash. It also includes built-in workflows such as planning and stress-testing tasks, and Meta says the agent can delegate work to sub-agents in parallel. TechCrunch reported that Meta is positioning the tool as a lower-cost alternative to OpenAI’s Codex and Anthropic’s Claude Code, though analysts have cautioned that coding benchmarks are not always directly comparable across products.
The launch arrives at a sensitive moment for Meta, which has been under pressure to show that heavy AI investment can translate into revenue. Foreign Policy Journal reported that Meta’s shares fell in the week before the launch after Mark Zuckerberg gave investors limited detail on cloud-computing plans. The same report said Muse Code placed second on the Terminal-Bench 2.1 benchmark, behind Anthropic’s Claude Code Opus 5 but ahead of OpenAI’s Codex, leaving Meta with a performance gap even as it competes aggressively on price. The launch was also overshadowed by a separate disclosure involving Meta’s Muse Spark 1.1 model, after a third-party evaluator misconfigured testing and temporarily granted unintended internet access, an episode both sides described as contained.
Disclaimer: This content is intended for informational purposes only. Readers are advised to exercise their own judgement, conduct due diligence, or consult a qualified expert before acting on any information provided.
