Muse Spark 1.2 vs. Muse Code

Muse Spark 1.2 is a coding-focused AI model from Meta designed to support advanced software engineering tasks through Muse Code and the Meta Model API. The model builds on Muse Spark 1.1 with improvements in code generation, complex debugging, codebase understanding, and end-to-end developer workflows. Muse Spark 1.2 powers Muse Code, a terminal coding agent that can plan repository changes, write code, validate outputs, and work across large codebases. Muse Code uses persistent async background agents that stay active throughout a session to reduce redundant information gathering and support difficult multi-step work. The runtime uses a local event log where model calls, tool runs, approvals, and edits are appended, making sessions replay-exact and restart-safe. Muse Spark 1.2 was co-trained with Muse Code so the model can take advantage of its toolset, harness workflows, goals, compaction, and subagent architecture. Meta significantly scaled training compute on coding tasks and expanded training environment diversity to improve the model’s engineering capabilities. The model was also trained on long-horizon coding tasks, including whole-repository generation, large end-to-end projects, auto-research, and extended iterative work. Its training approach uses planning, goal conditioning, context compaction, rejection-sampled harness trajectories, and self-improvement data generated with Muse Spark 1.1. Meta also tested Muse Spark 1.2 on long-running GPU kernel optimization workflows where the model wrote, compiled, profiled, and improved Triton kernels over many tool calls. By combining coding-focused training, agentic runtime integration, persistent subagents, long-horizon reasoning, replay-safe execution, and API availability, Muse Spark 1.2 helps developers and AI agents complete complex software engineering work with less intervention.

Muse Code is Meta’s terminal-based AI coding agent built to take on complex software engineering tasks across large repositories. The agent is powered by Muse Spark 1.2 and is designed to plan code changes, write implementation code, validate results, and support end-to-end developer workflows. Muse Code can coordinate multiple persistent subagents for each task, helping solve difficult problems faster and with less manual intervention. Its architecture uses a simple agent loop enhanced by async background agents that remain active for the full session instead of being spawned only for individual steps. These background agents reduce repeated information gathering, carry out next actions, and communicate back to the main agent when useful. Muse Code’s runtime uses a local event log where model calls, tool runs, approvals, and edits are continuously appended. This event log serves as a single source of truth, making the runtime replay-exact and restart-safe if a crash or interruption occurs. The design allows Muse Code to handle long-running development work without losing progress or context. Muse Code includes bundled skills such as /plan for approval-gated task planning, /grill for stress-testing plans, and /goal for working toward successful completion of a defined objective. Example workflows include interpreting a video input, understanding the requested output, and producing a rich software experience such as a vacation home marketing and booking page. By combining terminal execution, autonomous planning, persistent background agents, replay-safe runtime design, bundled skills, and Muse Spark 1.2 model support, Muse Code helps developers complete ambitious coding tasks with greater reliability.