Meta launched Muse Spark 1.2 on August 5, 2026, a flagship coding model built for complex software engineering tasks. The model ships with a one-million token context length, improved first-attempt coding accuracy, and deep integration with persistent asynchronous background agents.
Muse Spark 1.2 is purpose-built for debugging sprawling codebases, validating changes across thousands of files, and handling multi-step reasoning tasks. The one-million token context window means the model can hold the full picture of a project in its working memory rather than operating on isolated code snippets.
Meta is also launching Muse Code in beta alongside the model refresh. The tool targets macOS and Linux platforms and is designed to streamline developer workflows around debugging, code validation, and iterative development.
Muse Spark 1.2 is available exclusively through the Meta Model API, with pricing starting at $1.25 per million input tokens.
The predecessor to the current release, Muse Spark 1.1, launched on July 9, 2026, and established the initial API access and multimodal capabilities that Spark 1.2 now builds upon.
Contrary to earlier strategies promoting open weights in its Llama series, Meta is adopting a closed, API-centric model for its Muse lineup. There is currently no publicly verifiable information regarding an associated model called “Muse Glimmer,” nor any confirmed open-weight downloads for the Muse models, underscoring a shift in how Meta is managing commercialization of its AI developments.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.

1 hour ago
8









English (US) ·