Documentation

Muse Spark documentation

What the model accepts, how it plans, and how your subscription and credits map to generation. Written for the people who are about to put it to work.

Introduction

Muse Spark reads images, charts and text together, plans with multiple agents, and turns a single prompt into images, websites and mini-games. Every request starts in the composer on the home page: attach what you have, describe the outcome, and press Act.

Capabilities

  • Multimodal understanding. Text, images, charts, diagrams and tables are read in a single context.
  • Parallel reasoning. Contemplating mode runs several agents before reconciling an answer.
  • Agentic creation. Images, landing pages, dashboards and mini-games are produced from the plan.
  • Personalised recommendations. The same model class surfaces recommendations across Meta apps.

Reasoning modes

Standard

One pass: read, reason, answer. Best for summarising, rewriting, explaining a screenshot and answering questions about an attached document.

Contemplating

Multiple agents reason in parallel and the results are merged before the response is written. Use it for multi-step STEM and research questions, comparative analysis, and anything where an assumption should surface rather than hide.

Plans and credits

Generation is gated by an active subscription. Each plan grants a monthly credit allowance; annual plans grant a larger allowance at a lower effective monthly price. Credits are spent when a request is produced, and the balance is always visible from the account menu.

Sign-in runs through Google only. If you sign in without an active plan, the composer sends you to the pricing page — nothing is charged until you choose a plan in Stripe Checkout.

Support

Questions, billing issues and refund requests: support@musespark.homes.

Muse Spark documentation — capabilities, modes and plans