Meta launched Muse Spark 1.3, the latest version of its Muse family of AI models, with upgrades focused on coding and agentic tasks. The tech giant says the new model can handle longer workflows, use tools more effectively, and work across multiple tasks in a single conversation.
The launch comes as Meta intensifies its competition with AI leaders such as OpenAI and Anthropic. The companies are increasingly shifting their focus from chatbots that generate responses to AI agents that can plan and execute complex tasks.
Muse Spark 1.3 can sustain longer-horizon work. The model can work with messy or conflicting information, create context using available tools, and identify gaps in its own plans.
It can also manage several workflows within a single long conversation. This allows users to give the model broader objectives rather than breaking every task into individual prompts.
Meta says, “The model also improved its judgement around actions that could have irreversible consequences. It can assess such situations more carefully before proceeding.” Coding is one of the key areas of improvement in Muse Spark 1.3. The model performs better on longer-horizon coding tasks and says it now produces cleaner code while taking fewer unnecessary turns.
According to comparisons by Meta engineers, “Muse Spark 1.3 uses about 20% fewer tool calls and 25% fewer tokens than Muse Spark 1.2. This makes the model faster and more efficient in common software engineering workflows.”
Muse Spark 1.3 is rolling out through Muse Code and Meta Model API. Meta said its existing reasoning modes are available now, while the maximum reasoning mode will arrive after additional safety testing. The tech giant is also working on larger Muse models and an open-weights release, signalling that Muse Spark 1.3 is part of a wider effort to compete in the next phase of agentic AI.
Also Read: Meta Unveils Muse Glimmer to Bring Agentic AI to Personal Devices