While you read up on the Muse model family, generate your own images right here with the Muse Image model - free credits included.
Real generations from this workspace - click any card to reuse the prompt.
A proprietary agentic language model released in April 2026, updated to version 1.1 in July 2026. Highlights: a 1 million token context window, native multimodal reasoning, and strong agentic coding - it decides when to write a script and when to click through an interface.
Released July 7, 2026. An agentic image model that plans compositions before rendering, searches the web for factual details, and can even write code mid-generation to produce charts or working QR codes. Ranked #2 on the Arena text-to-image leaderboard within days of launch.
Previewed July 7, 2026 with native audio - one generation produces both picture and synchronized sound. It is still in early preview with no public release date, which is exactly why creators look for ways to make Muse-style videos today.
Muse Spark 1.1 reads entire codebases or document collections in one pass, with cached input priced at a fraction of fresh tokens.
The Model API launched at $1.25 per million input tokens and $4.25 per million output tokens - undercutting most frontier-model rivals at launch.
Reported scores include 88.1 on MCP Atlas and 61.5 on SWE-Bench Pro, positioning it as a serious agentic coding contender.
The LLM is available in consumer AI apps and through the public-beta Model API for developers, which includes free starter credit for new accounts.
The official image model is rolling out gradually by region and platform, with an unpublished free-use cap and no API. Availability varies by country.
This workspace gives you Muse-style image and video generation in any browser today - transparent credits, no waitlist, no region lock. Try it with the generator above.
Models plan before they render - reasoning about composition and facts instead of decoding a prompt straight to pixels.
Reinforcement learning taught the image model to review and fix its own drafts before showing you the result.
Charts, diagrams, and scannable QR codes can be generated as working artifacts, not just pictures of them.
Muse Video generates synchronized speech, ambience, and music together with the picture in a single pass.
The 1M token window lets the LLM hold an entire project in mind while it works.
Unlike the Llama era, the Muse family is proprietary - there are no open weights to self-host, which makes hosted access the only way in.
Context Window
1M
Tokens in Muse Spark 1.1
API Input Price
$1.25
Per million tokens
Arena T2I Rank
#2
Muse Image at launch week
Family Models
3
Spark, Image, and Video
Muse Spark is a proprietary agentic language model, first released in April 2026 and updated to 1.1 in July 2026. It succeeded the Llama series and marked the shift from open weights to a closed, hosted model strategy. The name also anchors the wider Muse family, which includes the Muse Image and Muse Video generation models.
At public beta launch, pricing was $1.25 per million input tokens, $4.25 per million output tokens, and $0.15 per million cached input tokens, with web search priced separately. New developer accounts received free starter credit.
Reported benchmarks include 61.5 on SWE-Bench Pro and 88.1 on MCP Atlas. It is positioned for agentic work - long multi-step tasks that mix writing code, running tools, and operating interfaces - rather than just single-file completions.
No. As of July 2026 the image model has no public API and no committed timeline for one. Only the Muse Spark LLM is available through the Model API.
No date has been announced. It was previewed on July 7, 2026 as an early preview with native audio, described as coming soon to creators. Until it ships, browser workspaces like this one are how people create Muse-style videos.
The LLM is broadly available via app and API. The image features roll out by region and platform, and notably have not launched in the EU. This workspace, by contrast, works worldwide.
Yes. The generator on this page uses the Muse Image model for stills, and the video workspace uses Muse Video for clips. New accounts get 55 free credits, and paid downloads carry no watermark.
No. Muse AI Image is an independent creative platform and is not affiliated with, endorsed by, or sponsored by Meta Platforms, Inc. This page summarizes publicly reported information about the Muse model family for reference.
