Official documentation describes the named model or gateway—not every feature available in this app.
Wonder’s API identity and reasoning controls
Meta lists muse-wonder-1.0 in preview with a 1,048,576-token context and minimal, low, medium, high, xhigh and max reasoning effort. It understands text, images, video, audio and PDFs and returns text. Omitted effort lets Meta choose reasoning depth; the Standard-only API model is distinct from Wonder Full.
Muse Spark shares context, not identical behavior
The model table lists a 1,048,576-token context for Spark 1.1, 1.2 and 1.3, with text, image, video, audio and PDF input and text output. The 1.3 audio caveat qualifies that table. Context capacity is separate from file ingestion limits and does not guarantee that every detail is recalled.
Reasoning effort and data tier are linked
Muse Spark does not accept reasoning_effort none. The max level is restricted to Standard-tier Muse Spark 1.3; it is not available on Contributor variants or the earlier Spark versions. Meta describes Contributor as a discount in exchange for permission to train on prompts and completions, so it is not just another speed setting.
Muse Image has distinct endpoint options
Meta documents Responses for conversational editing and separate images/generations and images/edits endpoints for single-shot work. The model is muse-image-1.0. These provider routes do not establish which tools, references or continuity settings the app forwards; the audited app route is described separately below.