Muse Spark 1.1

Released July 9, 2026 by Meta Superintelligence Labs. Muse Spark 1.1 is a multimodal agentic reasoning model: it accepts text, image, video, PDF and audio input, outputs text, and actively manages a 1M-token context. It is also the first Meta model ever offered as a paid API — a launch announced alongside Mark Zuckerberg's first X post in about three years.

Muse Spark 1.1 strengths

• Multimodal input: text, image, video, PDF and audio
• Actively managed 1M-token context
• Free in Thinking mode in the Meta AI app and on meta.ai (Meta login required)
• Developer access via the Meta Model API with $20 free starting credits

Pricing & context

Cost guide: a typical call of about 10K input + 2K output tokens costs roughly $0.021 at list prices. New Meta Model API accounts include $20 in free starting credits, and Thinking mode in the Meta AI app and on meta.ai is free with a Meta login — so you can evaluate before paying anything.

Meta-reported benchmark results

Meta reports MCP Atlas 88.1, Humanity's Last Exam with tools 62.1 and JobBench 54.7 for Muse Spark 1.1. Treat these as Meta-reported figures, not independent Benchquill test results.

When to choose Muse Spark 1.1

Muse Spark 1.1 is best for agentic, tool-using workloads that mix media types — it takes text, image, video, PDF and audio in a single 1M-token context and manages that context actively across long-running tasks. Note that the Meta Model API is currently a US-only public preview, so non-US teams are limited to the free Thinking mode in the Meta AI app and on meta.ai for now.

Muse Spark 1.1 FAQ

How much does Muse Spark 1.1 cost?

Muse Spark 1.1 is free to use in Thinking mode in the Meta AI app and on meta.ai (Meta login required). Via the Meta Model API — a US-only public preview — it costs $1.25 per 1M input tokens and $4.25 per 1M output tokens, with $20 in free starting credits. A typical 10K-input / 2K-output call costs roughly $0.021.

What is Muse Spark 1.1 best for?

Muse Spark 1.1 by Meta Superintelligence Labs is best for agentic, tool-using workloads that mix text, image, video, PDF and audio input inside an actively managed 1M-token context.

Is Muse Spark 1.1 multimodal?

Yes. Muse Spark 1.1 accepts text, image, video, PDF and audio input and outputs text.

How does Muse Spark 1.1 score on benchmarks?

Meta reports MCP Atlas 88.1, Humanity's Last Exam with tools 62.1 and JobBench 54.7 for Muse Spark 1.1. These are Meta-reported results, not independent Benchquill tests.

Tools that use Muse Spark 1.1