Muse Spark 1.2 is now available on Vercel AI Gateway

**Muse Spark 1.2** from Meta is now available on **Vercel AI Gateway**. It is a coding-focused update to the previous Muse Spark model. The release keeps the model’s general capabilities while targeting stronger performance on software work.

The update focuses on **code generation**, **complex debugging**, **codebase understanding**, and **end-to-end developer workflows**. It is built for long-horizon work such as generating whole repositories, building large projects end to end, and sustaining iterative development across multi-step tasks rather than single-shot completions.

For engineers and builders, that matters because many real tasks span more than one file or one turn. A model that can hold context across a repository, debug non-trivial failures, and carry a project through successive changes reduces handoffs between tools and shortens the loop from intent to working code.

Placing Muse Spark 1.2 on **AI Gateway** puts a Meta coding model behind the same access path Vercel already uses for other providers. Teams that already route models through the gateway can try 1.2 without a separate Meta integration path, and can compare it with other coding models under one routing and ops surface.

Practical next step: if you already use AI Gateway for coding agents or IDE-adjacent workflows, route a subset of repository-scale and multi-file debugging traffic to Muse Spark 1.2 and measure quality against your current default on the same tasks. Watch how well it sustains long-horizon sessions—whole-repo generation, large project build-out, and multi-turn iteration—rather than only single-prompt code snippets.