# Meta Ships Muse Code, Its First Coding Agent, on a New Muse Spark 1.2 Model

The terminal agent runs parallel subagents against large repositories and prices at $1.25 per million input tokens, with a cheaper tier for users who let Meta train on their prompts.

- Published: 2026-08-08T06:26:48.459Z
- Canonical: https://polylog.news/ai/2026-08-08/meta-ships-muse-code-its-first-coding-agent-on-a-new-muse-sp
- Publisher: Polylog (AI desk)
- Section: tech
- Sources: [Polylog editors](https://polylog.news), [Meta AI Developers Blog](https://developer.meta.com/ai/resources/blog/build-with-muse-code/), [VentureBeat](https://venturebeat.com/orchestration/meta-enters-the-ai-coding-wars-with-muse-spark-1-2-and-muse-code-with-persistent-async-background-agents), [The Register](https://www.theregister.com/ai-and-ml/2026/08/06/meta-wants-to-get-inside-your-terminal-with-its-new-coding-agent/5283717)

Meta released an early beta of Muse Code, a command-line coding agent, together with Muse Spark 1.2, a coding-focused model the company says it [co-trained with the agent](https://developer.meta.com/ai/resources/blog/build-with-muse-code/) so that model and harness fit each other. Meta describes gains over Muse Spark 1.1 in code generation, complex debugging, and codebase comprehension, which it attributes to more coding training compute and a wider set of training environments.

The agent's design choices are the more interesting part. Muse Code installs from the terminal with one command, runs several subagents in parallel with implementation and review happening at the same time, and keeps a replayable event log of every subagent, tool call, intervention, and cancellation. [VentureBeat reported](https://venturebeat.com/orchestration/meta-enters-the-ai-coding-wars-with-muse-spark-1-2-and-muse-code-with-persistent-async-background-agents) persistent asynchronous background agents as the headline capability, and coverage put the model's context window at one million tokens. That is an argument for auditability, directed at engineering managers who will not let an unlogged agent touch a production repository.

Pricing is where Meta is doing something other labs have avoided stating explicitly. The standard muse-spark-1.2 rate on the Meta Model API is $0.15 per million cached input tokens, $1.25 per million input tokens, and $4.25 per million output tokens. Alongside it sits a "contributor" tier at a materially lower price in exchange for Meta using the user's prompts and completions as training data. That converts a privacy preference into an explicit line item.

No independent benchmark reproduction has appeared yet. Meta's claims of improvement over version 1.1 are internal, and the comparisons that matter, against Anthropic's Claude Code and OpenAI's coding stack on tasks like real repository-level pull requests, have not been published by a neutral party. [The Register noted](https://www.theregister.com/ai-and-ml/2026/08/06/meta-wants-to-get-inside-your-terminal-with-its-new-coding-agent/5283717) that Meta is entering a category where the incumbents already have developer habit on their side.

## What this means

Meta is competing on distribution and price rather than on a claimed capability lead, and the contributor tier is the mechanism: it buys training data with a discount instead of paying for it. That pressures Anthropic and OpenAI on the low end of coding-agent pricing, and it gives enterprises with strict data-handling rules a reason to stay on the standard tier at rates comparable to rivals. The replayable event log targets the same buyer, since audit trails are what unblock agent deployment in regulated engineering organizations.

## What to watch

- Whether independent benchmarks on repository-level coding tasks put Muse Spark 1.2 near the leaders or well behind, which decides whether the price tier is a bargain or a discount on weaker work.
- How many teams take the contributor tier. Heavy uptake would show that developers will trade code-adjacent data for cost, and would give Meta a training data stream its rivals must pay for.
- Whether Anthropic or OpenAI responds with a matching low-cost coding tier, which would confirm that agent pricing, not model quality alone, is now the competitive front.