Meta has launched Muse Spark 1.1, a new multimodal reasoning model designed for coding, computer control, tool use, and complex agent-style work.
It sounds powerful. It also sounds like something most consumers will go out of their way to avoid.
According to Meta, Muse Spark 1.1 is a substantial upgrade over the original model. It can work across multiple apps, understand images and video, write code, operate computers, call external tools, and manage long projects with less human involvement.
The model also has a 1 million-token context window, because no modern AI announcement is complete without a context-window number large enough to make everyone stop asking what people will actually use it for.
Meta says Muse Spark 1.1 can plan a job, divide it among several subagents, keep track of earlier decisions, and adjust when new information appears. It can also decide whether a task is better handled by clicking through an interface or writing a script.
One demonstration shows the model turning a smartphone video into a Facebook Marketplace listing. It can select photos, identify the item, write the description, and operate the browser on the user’s behalf.
That may be convenient. It may also sound like a nightmare to anyone who does not want Meta analyzing personal videos and taking control of a browser.
Developers can access Muse Spark 1.1 through the new Meta Model API, which is launching in public preview. The model is also available in Thinking mode through the Meta AI app and Meta’s website.
Meta included the usual collection of charts, benchmarks, and friendly quotes from partners. Replit praised the model’s coding, multimodal support, tool calling, and structured output. Box said it performed competitively on enterprise workloads.
Maybe it did.
These are still claims and endorsements selected for Meta’s own launch announcement. Independent testing will tell us whether Muse Spark 1.1 is truly competitive or merely another model that looks fantastic in a company-controlled presentation.
Even if it is excellent, Meta has a much bigger problem.
A lot of people simply do not trust the company. They do not trust Facebook. They do not trust Mark Zuckerberg. They certainly do not seem eager to hand Meta even more personal information while allowing its AI to watch videos, inspect documents, browse websites, and perform actions on their behalf.
Meta keeps acting as though better AI performance will automatically create demand. That is not how this works.
Consumers already appear tired of AI being stuffed into every app and service. Meta AI has been pushed across Facebook, Instagram, Messenger, and WhatsApp, but being impossible to avoid is not the same thing as being popular.
Muse Spark 1.1 could be fast, capable, and genuinely useful. None of that matters if people see the Meta logo and immediately look for the off switch.
Meta may have built another impressive AI model. Its real challenge is convincing anyone outside its own ecosystem to care.
Support independent tech journalism
NERDS.xyz is independently owned and operated. If you enjoy my coverage of Linux, AI, hardware, cybersecurity, and tech culture, consider supporting the site on Ko-fi.
Support NERDS.xyz