Meta launches Muse Image, native AI image model to reduce third-party dependence

Meta has released Muse Image, its first in-house AI image-generation model, marking a shift away from reliance on third-party tools like Midjourney and Black Forest Labs. Originally codenamed Mango, Muse Image is the second major release from Meta Superintelligence Labs under Alexandr Wang, following the April debut of Muse Spark, the company's next-generation LLM that replaced the Llama family.

The model will be available free to consumers via Meta AI, WhatsApp direct messages, and Instagram Stories, with power users and creators accessing a premium tier through Meta One subscriptions (introduced in May). Advertisers gain access through Meta's Advantage Plus service, where Muse Image powers on-brand ad-creative generation with fewer iterations. Internal benchmarks show Muse Image trailing OpenAI's GPT Image 2 but outperforming Google's Nano Banana 2 on single- and multi-image editing tasks.

Meta's strategy signals both a cost-reduction play (avoiding recurring payments to external vendors) and a consumer-retention move to compete with OpenAI and Google's image-gen offerings. The company plans a Muse Video model for later release. For AI practitioners and advertisers, the availability of a native image-gen tool in the Meta ecosystem lowers friction for integrated workflows and reduces reliance on multi-vendor stacks.