Muse Spark 1.3 is trained for agentic workflows and optimized for competitive coding performance. Developers can expect higher first-attempt accuracy and reliable tool calling. It tracks context and prior results, works through messy or conflicting inputs, and asks for input when needed.

Tuned for long-horizon coding workflows, it performs competitively with frontier models across several coding evals. Native multimodal perception allows Muse Spark to perceive video, images, and documents, with visual reasoning running through a real execution environment. The model supports screenshots and clips as input for building software.

Pricing includes two tiers: muse-spark-1.3-contributor at $0.10 input, $0.002 cached input, $0.20 output per Mtok, and muse-spark-1.3 at $1.25 input, $0.15 cached input, $4.25 output per Mtok. Resources include quickstarts, multi-agent orchestration, agent fan-out, computer use, GitHub agent, and search grounding cookbooks. Deployment is via Meta Model API.

Meta Products terms and cookie policies apply, with essential and optional cookies detailed for user control.