Meta: Introducing Muse Code and Muse Spark 1.2

Meta AI:

We’re excited to release Muse Code (beta), a terminal coding agent powered by Muse Spark 1.2, our newest model. This marks our next step toward the frontier, with larger and much more capable models on the way. [...]

Muse Code takes on complex software engineering tasks across large repositories: planning changes, writing code, and validating the results. It can coordinate multiple persistent subagents for each task, solving difficult problems faster, more accurately, and with less intervention.

Terminal-only (so far), so it avoids the whole native-app-versus-Electron-shit-sandwich debate. They’ve embedded some fun interactive examples of Spark’s output right in the web page.

Simon Willison:

An interesting twist on pricing is that the model is offered as two different model IDs. muse-spark-1.2 is priced at $1.25/million input and $4.25/million output — close to Gemini 3.6 Flash ($1.50/$7.50) — but if you agree to let Meta use your data “to improve our products” you can use muse-spark-1.2-contributor which is $0.10/$0.20 — a huge discount, closer to GPT-5.6 Luna ($0.20/$1.20) and Gemini 3.1 Flash-Lite ($0.25/$1.50).

That’s over a 10× discount in exchange for giving your data to Meta. I can see the appeal if you don’t care about the privacy of your data and code, like, say, if you’re using Muse Spark to generate code for a throwaway project, or for something that’s open source anyway. But I wonder if Meta has considered that this offer might spook would-be users who absolutely do not want to grant Meta the rights to their data. One would hope that the two systems have a strong firewall between them. But given Meta’s well-established contempt for the sanctity of user data, it’s not at all unreasonable to suspect that the difference here is like reserving a no-smoking seat in an airplane with a smoking section.

★ Friday, 7 August 2026