← Models

Muse Image

Meta · Muse

Meta's agentic image model for generation, editing, multi-reference composition, and tool-assisted refinement.

Type
image
Context
N/A
Max Output
N/A
Status
current
API Access
No
License
proprietary
image-generation image-editing multi-reference tool-use self-refinement social-context watermarking
Released July 2026 · Updated July 10, 2026

Overview

Freshness note: Model capabilities, limits, pricing, and regional availability can change quickly. This profile is a point-in-time snapshot last verified on July 10, 2026.

Muse Image is Meta’s current image-generation model from Meta Superintelligence Labs. It launched on July 7, 2026 for image generation, precise editing, and multi-reference composition inside Meta AI and selected Meta social surfaces.

The model is notable because Meta presents it as an agent rather than a one-pass prompt-to-image system. Muse Image can use search and code tools, inspect and refine its own generations, and plan jointly with Muse Spark. That makes it part of Meta’s broader assistant stack rather than a separate creator-only endpoint.

Capabilities

Muse Image supports text-to-image generation, instruction-based editing, iterative refinement across turns, and composition from multiple reference images. References can contribute people, objects, clothing, styles, and environments, with text and images interleaved inside more complex prompts.

Meta also documents three agentic behaviors:

  • Search can provide current factual and visual references for knowledge-heavy prompts.
  • Code execution can improve precise elements such as plots, figures, and QR codes.
  • Self-refinement can locally edit a draft, restart a generation, or choose a different tool-assisted approach when the first result is weak.

Meta says additional inference-time reasoning, tool calls, and refinement steps can improve human-preference results. Treat those as provider findings and test the model on your own brand, typography, factuality, and identity-preservation requirements.

Technical Details

Muse Image is an image-native model, so its displayed token-style context and maximum-output fields use 0 to mean N/A. Meta has not published a public model ID, downloadable weights, a self-serve API contract, or token-based rate card.

Images created through the Meta AI app and meta.ai carry Content Seal, Meta’s invisible provenance watermark. Meta says the signal is designed to survive common transformations such as cropping, compression, resizing, and screenshots, and it is previewing a detector for checking whether an image carries that signal.

Pricing & Access

Official launch availability:

  • Meta AI app and meta.ai
  • Instagram Stories in the US
  • WhatsApp in limited countries
  • Facebook planned for a later rollout

Meta has not documented public API access or API pricing. Consumer availability and creation limits can vary by country, account, Meta product, and subscription, so production teams should not treat the consumer launch as a stable developer endpoint.

Best Use Cases

Muse Image fits iterative social visuals, multi-reference compositions, small-business campaign concepts, factual graphics that benefit from search or code, and Meta-native creative workflows. It is especially relevant when the surrounding experience already uses Meta AI or Muse Spark.

It is less suitable when you need a public API, predictable per-image pricing, deterministic automation, or deployment outside Meta’s product surfaces.

Comparisons

  • GPT Image 2 (OpenAI): Better documented as an API model; Muse Image is more tightly integrated with Meta AI and social context.
  • Nano Banana Pro (Google): Competing image generation and editing route inside Google’s Gemini ecosystem.
  • Grok Imagine Quality Mode (xAI): Public API-oriented image model with explicit pricing; Muse Image currently remains product-first.