Meta is putting its Muse Spark 1.1 model in front of developers through a public preview of the Meta Model API, giving US developers a way to connect the model to their own tools instead of using it only inside Meta’s consumer AI surfaces.
The company says the 1.1 release is a major jump from the first Muse Spark model it introduced in April. Meta attributes the changes to developer feedback, and is pitching the update at software work in particular: more advanced coding, finding and fixing complex bugs, and support for workflows that rely on multiple AI agents coordinating across apps.
That is Meta’s claim, not an independent benchmark. The company did not provide comparative test results in the announcement described by The Verge, so the “ready to compete” part remains marketing until developers run it through real projects, ugly repos, weird dependencies and all the other things demos tend to avoid.
What developers get
Muse Spark 1.1 is available now in Thinking mode through the Meta AI app and the Meta AI website. The more developer-relevant change is the Meta Model API, which Meta is opening in public preview for US developers.
API access means developers can call the model from coding assistants, agent systems or other applications rather than asking users to stay inside Meta AI. Meta says every new Meta Model API account gets $20 in free credits.
Meta also says Muse Spark 1.1 has native multimodal perception for images, videos and documents. In plain terms, the company says the model can work across those input types without requiring a separate bolt-on system for each one. The announcement does not include enough detail to assess how well that works in practice.
Meta’s broader AI catch-up plan
The launch follows Meta’s release earlier this week of Muse Image, an image-generation model that drew criticism because it can incorporate other Instagram users’ content into generated images, according to The Verge.
Meta has been trying to prove that its AI spending can produce products that stand beside systems from OpenAI, Google and Anthropic. The Verge reported that push comes after billions of dollars in investment, high-profile hiring and a company restructuring last year.
Muse Spark’s distribution has widened quickly. The first model was initially available directly through Meta AI. It later powered chatbots in Instagram and WhatsApp, as well as Meta’s latest smart glasses, according to The Verge.
The developer API is the more consequential step for people building software. Consumer chatbots can hide weak spots behind guardrails and interface choices. APIs expose models to customers who will measure latency, cost, coding accuracy and failure modes because those things break products. Meta now has to let developers kick the tires, which is where the polite launch language usually starts meeting the stack traces.
This story draws on original reporting from The Verge.
