What Meta Muse Code actually launched with—and what remains unverified

Meta Muse Code’s confirmed launch was a beta release on August 5, 2026, powered by the Muse Spark 1.2 model. The supplied reporting describes a terminal-based coding agent with usage-based access—not a general-availability product with $5 monthly subscriptions. 123

That distinction matters because the reported follow-up changes—general availability, monthly plans, request limits of 10 to 50 per five-hour window, and the retirement of the contributor tier—are not independently established by the sources provided here. They should be treated as unverified rather than presented as completed product changes.

What Muse Code was built to do

Muse Code is intended to handle software-engineering work across large repositories rather than only generate isolated snippets. Meta describes a workflow that covers planning changes, writing code, and validating the results. Muse Spark 1.2 was developed as a coding-focused update with improvements in code generation, debugging, codebase understanding, and end-to-end developer workflows. 23

The launch reporting also says Muse Code can work on long and complex coding projects while running multiple sub-agents simultaneously. 1 Meta’s product design pairs a primary agent with background workers that can gather information, carry out steps, and report their findings, reducing the need for developers to manually direct every stage of a multi-step task.

The agent also emphasizes visibility into its work. Reporting on the beta describes a local event log that records model calls, tool use, approvals, and edits, allowing developers to inspect activity and resume work after an interruption. The interface includes commands such as /plan, /grill, and /goal for planning, challenging a plan, and pursuing an objective. 14

The confirmed pricing model at launch

At launch, developers could use Muse Code through a pay-as-you-go model tied to Muse Spark 1.2. The standard listed rates were $1.25 per million input tokens and $4.25 per million output tokens. 317

The sources also describe a lower-cost contributor tier. Its reported rates were approximately $0.10 per million input tokens and $0.20 per million output tokens, in exchange for allowing prompts and completions to be used to improve Meta’s products or models. 71826

That contributor option was not simply a cheaper unlimited plan. The available descriptions indicate that it was subject to token-based limits over a rolling five-hour period, while the standard tier used usage-based API pricing. 1823 The exact terms and availability should be checked against Meta’s current documentation before developers submit proprietary code or prompts.

What the reported subscription change would mean—if confirmed

A move from beta, metered API access to general availability and fixed monthly plans would represent a meaningful product shift. It would make costs easier for individual developers and small teams to forecast, while positioning Muse Code as an end-user developer product rather than only a way to access Meta’s model infrastructure.

It would also place Meta more directly in the terminal-agent competition involving Anthropic’s Claude Code and OpenAI’s Codex. Early coverage already framed Muse Code as Meta’s entry into that market. 417 In this category, the practical comparison is not just model quality. Developers also need to consider repository-scale reliability, tool integrations, approval controls, background-agent behavior, speed, privacy terms, and how quickly usage limits are consumed.

But none of that confirms that the reported $5 plans exist. The supplied evidence supports the original token prices and beta status; it does not establish whether token billing was replaced, retained for API access, or supplemented by subscriptions.

Why fixed monthly plans would be difficult to design

Coding agents can consume substantially more compute than a single conversational answer. A task may involve repository exploration, several tool calls, code edits, tests, review passes, and parallel workers. A fixed monthly price would therefore need meaningful boundaries around usage.

Those boundaries could include rolling request windows, task-length limits, concurrency restrictions, priority levels, or paid overages. As a result, a low advertised monthly price would not by itself show that Muse Code offers better value than token billing or competing agents. The important question would be how much useful repository work each tier actually supports before a user reaches its limit.

The questions developers should still ask

Is there evidence of real-world adoption?

The supplied sources establish what Meta says Muse Code can do, but they do not provide independent evidence about production-repository reliability, task-completion rates, developer retention, or preference versus Claude Code and Codex. Those metrics remain open.

Did Meta really replace token pricing?

The confirmed launch pricing was $1.25 per million input tokens and $4.25 per million output tokens. 317 The available evidence does not show whether a later subscription announcement superseded those rates or created a separate product tier.

What happened to the contributor tier?

The contributor model was described as substantially cheaper but tied to data-use permissions and rolling limits. 71826 The supplied evidence does not reliably establish that Meta ended the option, changed its discount, or altered what happened to data previously submitted under its terms.

Are the reported limits request-based or token-based?

This is a material distinction. The launch sources describe contributor limits in terms of tokens over a rolling five-hour window, not the reported 10–50 requests per five-hour structure. 1823 Without an authoritative product announcement or current pricing page confirming the newer figures, the request caps should not be treated as established fact.

Bottom line

The evidence supports a clear launch story: Meta released Muse Code in beta on August 5, 2026, paired it with Muse Spark 1.2, and offered terminal-based software-engineering workflows with usage-based pricing. 123

The claimed Monday transition to general availability and $5 monthly subscriptions may describe a later development, but it is not verified by the supplied sources. For now, the defensible conclusion is that Meta had entered the competitive coding-agent market with a beta product—and that its pricing, contributor data policy, adoption, and long-term subscription strategy still require confirmation.