Unverified50% confidenceFactExact time
Muse Spark Open has a built-in 1.8-billion-parameter dedicated perception encoder for processing images and video frames, audio is handled by a separate encoder, and inputs from all modalities are ultimately converged into text output.
1
Sources
50%
Confidence
Long-term
Relevance
10/16/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedMuse Spark Open supports multimodal input of text, images, video and audio.80% similarUnverifiedMuse Glimmer has multimodal capabilities, with a built-in vision encoder that can accept both text and image inputs simultaneously77% similarUnverifiedThe Muse Wonder model card lists a 1-million-token context and support for text, image, video, and audio input, with text output.77% similarUnverifiedMuse Glimmer adopts a 52-layer architecture, supports text and image input, and covers more than 100 languages.68% similarUnverifiedMeta Muse Spark Open's input modalities cover text, images, video and audio, and the degree of reasoning effort can be adjusted between low and xhigh.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1047113API
curl https://kongchang.com/api/v1/knowledge/claims/1047113MCP
get_claim(id=1047113)