Meta launches agentic Muse Image and Video models for AI content generation

Meta Superintelligence Labs has released two new generative models, Muse Image and Muse Video, aiming to integrate agentic capabilities directly into content creation. Muse Image, designed for generating and editing images, reportedly achieved top rankings on the LMArena leaderboard shortly after its July 7, 2026 launch. Unlike traditional models, Muse Image can reportedly execute code and perform web searches to verify the accuracy of its outputs, such as ensuring correct data representation in charts or functional QR codes. Muse Video is currently in early preview with known issues, while Muse Image also boasts precise editing and multi-reference composition features. AI

IMPACT These agentic models could set a new standard for accuracy and verification in AI-generated content, potentially impacting workflows requiring precise visual data.

RANK_REASON Frontier-lab model release with system card and benchmark claims. [lever_c_demoted from frontier_release: ic=1 ai=1.0]

Read on dev.to — LLM tag →

• ByteDance
• Claude
• DeepSeek
• Gemini
• generative pre-trained transformer
• Google
• Meta
• Meta AI
• Muse Image
• Muse Video
• Nano Banana
• Qwen
• See Dream

• model release
• product

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →