Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
Dead on Thursday Next version announced Of Spark AI’s multimedia model, Muse Spark 1.1, describing it as A big upgrade from its predecessor. The model is designed specifically for agentive tasks and promises significant gains in areas such as computer use, coding, and multimedia understanding.
Spark 1.1 comes after the launch of Image of Mosesthe new photo creation tool, which some Instagram users have already opted out of. (You can learn How to do it here.)
With proverbs OpenAI and Anthropic Meta’s announcement, which makes headlines with new, advanced AI models, is a signal that the company is getting back into the AI race after falling behind for more than a year and undergoing an extensive reorganization. The model was released through what Meta calls Superintelligence Labs, a group led by former Scale AI CEO Alexander Wang.
Meta has also launched a public preview of its Meta Model API, where the new Spark 1.1 will be available to developers. The model is now available in “Thinking” mode within the Meta AI application and beyond meta.ai.
Enhancing agent capabilities
Meta says Spark 1.1 can complete complex projects much faster than its predecessor. He’s trained to “coordinate multi-agent systems to improve end-to-end latency,” according to a company blog. Agent AI refers to models that can take actual actions on your behalf, rather than just providing answers to a prompt.
This multi-agent system features a master agent and sub-agents. The master agent will develop a plan and delegate its implementation via parallel sub-agents. These sub-agents will stick to their assigned tasks, use available tools and escalate back to the master agent when necessary.
Spark 1.1 also effectively manages the million-token context window by remembering actions, retrieving information from previous work, and compressing context in a way that preserves important steps needed for later, according to the blog. This means that you should be able to ask it to do more complex tasks without falling into the limits of how much the model can handle at one time.
Computer use and coding
The new model may also be able to Take control of your computer For specific tasks, if that’s something you want to try.
Meta says its new AI model has received a boost in computer use workflows across multiple applications. It can maintain context across long sessions and can navigate unfamiliar interfaces with minimal user interaction. It can understand when certain tasks are automated and when the computer is navigated manually.
On the programming front, Spark 1.1 promises major improvements, especially in tasks involving complex databases. It can diagnose and fix errors, add new features in enterprise-level systems, and perform large code migrations.
Truly multimedia
Spark 1.1’s multimedia capabilities allow tasks to be seen and heard to be completed. This is not unlike the current AI models we’ve seen today, like the Gemini Live camera experience. It can ingest images and audio, and provide “highly descriptive” captions from sources. By combining this with his computer abilities, he can interpret what appears on the screen and take actions to carry out tasks.
safety
Meta touts comprehensive safety evaluations of the model, which follows an advanced AI measurement framework. The model is said to show strong resistance to jailbreaking, flash injection, and other common attacks.
