$ cat wiki/models/muse-image.md
Muse Image
modelupdated 2026-07-20created 2026-07-20
Spec
| Attribute | Value |
|---|---|
| Developer | Meta AI (Meta Superintelligence Labs) |
| Released | 2026-07-07 |
| Announced | 2026-07-07 |
| Context window | unknown |
| Pricing | Free (consumer apps); API pricing unknown |
| License | proprietary |
| Availability | Meta AI app, meta.ai, Instagram Stories (US), WhatsApp (limited) |
Release Date
July 7, 2026.
Benchmarks
- Arena text-to-image: #2 (human-preference Elo at launch)
- Arena single-image editing: #2
- Arena multi-image editing: #2
Use Cases
- Text-to-image generation
- Image editing (single reference)
- Multi-image composition
- Agentic generation with search + code tool invocation
Architecture / Key Features
- Agentic: invokes search and coding tools to improve accuracy; self-refines outputs
- Test-time compute scaling: generation quality improves with more inference compute
- Content Seal: invisible AI watermarking system to verify AI-generated images
- Integrates with Muse Spark (1.0 / 1.1) (cross-modal reasoning)
Compared To
| Model | Arena rank | Agentic? |
|---|---|---|
| Muse Image | #2 (text-to-image, editing) | Yes |
| Grok Imagine Video 1.5 (Preview) (predecessor comparison) | #1 at Video Arena | No |
| Stable Diffusion family | varies | No |
Sources
- Meta AI blog: Introducing Muse Image and Muse Video
- Facebook blog: Muse Image
- CNBC: Meta debuts Muse Image