Meta Superintelligence Labs launched Muse Image on July 7, its first in-house image generation model. The rollout comes with an early preview of Muse Video and drops the product straight into Meta AI, meta.ai, Instagram Stories in the US and WhatsApp in a limited set of countries.
Key Takeaways
- Muse Image is the first image generator built by Meta Superintelligence Labs, live from day one across Meta AI, meta.ai, Instagram Stories US and WhatsApp.
- On the Arena Elo leaderboard as of July 5, 2026, Muse Image ranks #2 for text-to-image and #2 for both single-image and multi-image editing.
- A preview of Muse Video ships alongside the launch, ranked #3 on Arena Elo text-to-video, with wider access for creators and Meta AI announced as coming soon.
Have an AI Sum Up This Article
ChatGPTMeta Superintelligence Labs ships its first image model
Meta Superintelligence Labs unveiled Muse Image on July 7, 2026 as its first in-house image generation model. The lab, built to put Meta back in the frontier race after years of ceding ground to OpenAI and Anthropic, chose to lead with image rather than text.
The model does more than map prompts to pixels. Meta describes it as operating as an agent: Muse Image invokes search and coding tools to improve accuracy, self-refines its own generations mid-run, and scales quality by burning more compute at inference time.
Muse Image also composes from multiple reference images and taps into Instagram social context. Creators can @-mention a public Instagram account to pull visual inspiration straight from that feed, without leaving the app or exporting anything manually.
The launch drops the product directly into Meta AI and meta.ai, into Instagram Stories in the US, and into WhatsApp in a limited set of countries. Meta laid out the use cases and rollout in its official Muse Image announcement published Tuesday. Facebook is flagged as coming soon with no public timeline. Meta AI on Instagram has had rougher moments, a hack hijacking the chatbot to steal accounts.
A #2 Arena Elo ranking and a video preview
On the public Arena Elo table as of July 5, 2026, Muse Image ranks #2 in text-to-image generation, #2 in single-image editing, and #2 in multi-image editing. Meta does not publish a direct numerical comparison against any named competing model in its launch post.
Muse Video ships alongside the launch as an early preview. The video model ranks #3 in text-to-video on Arena Elo on the same date, with a product pitch that leans on visual fidelity, temporal consistency and native audio support.
The video preview is capped for now to Meta’s cited “creators” cohort and to Meta AI itself, with no general access date. Muse Video sits behind a controlled beta while the image model already runs at scale for everyday users.
Also on Horizon:
- Meta Muse Image Now Lives in Instagram and WhatsApp
- DeepSeek Builds Its Own AI Chip to Cut Nvidia Reliance
- iOS 27 Finally Lets You Reshape Siri’s Voice
What it changes on product and competition
Short term, Meta plugs a missing piece of the stack. The group had been running image generation on third-party models with a cost-performance trade-off. An in-house model puts the lever back in Meta’s hands and cuts vendor dependency for billions of daily users.
Instagram turns into an inspiration source usable directly inside the product. A creator can ask Muse Image for an image inspired by the style of a public account with a single @-mention. The full “inspire then generate” chain now runs end-to-end inside Meta, with no external tool or manual export in the middle.
Medium term, the product-plus-model loop echoes what Google is doing with Gemini and Nano Banana: every major ecosystem now pushes its own image model inside its legacy apps. Pressure on pure-play image players like Midjourney, Ideogram and Runway rises another notch when a “good enough” model runs for free inside the Instagram feed.
On video, Muse Video shows up in a still-open category. The #3 Arena Elo ranking suggests a competitive but not dominant model. The real signal will come from the undated general opening and from standardized video benchmarks as the community actually gets its hands on the preview. Muse Image wins round one, Muse Video plays round two on its own clock.
Follow the story on Horizon.
