Meta releases Muse Spark 1.3 with stronger agentic and coding capabilities

MUMBAI: Meta is giving its AI agents another spark, releasing Muse Spark 1.3 with improvements aimed at long-running agentic workflows, coding and instruction-following.

The company announced the model on September 2, saying it had incorporated lessons from the wider adoption of Muse Code and Meta Model API to make Muse Spark 1.3 more useful in real-world applications.

Muse Spark 1.3 is rolling out immediately through Muse Code and Meta Model API. Existing reasoning modes are available, while the maximum reasoning mode is expected to follow after additional safety testing.

A key focus of the new model is its ability to handle longer and more complicated workflows. Muse Spark 1.3 can work across multiple tasks within a single conversation, use tools to build context from messy or conflicting information, identify gaps in its plans and retain what it has learned while working towards a final output.

Meta said the model has also been trained to collaborate more actively with users. It can ask clarifying questions when instructions are unclear, seek user input when it gets stuck and request confirmation before carrying out consequential actions.

For lengthy assignments, the model can adapt its working style to user preferences, either providing regular updates or continuing more quietly in the background.

Muse Spark 1.3 is also designed to follow complex, long-form instructions more reliably. Meta said it is better at retaining detailed requirements across multi-step tasks without dropping constraints or drifting from the requested workflow.

The model has received upgrades to its multitasking capabilities as well. It can more accurately identify which task an incoming prompt relates to in a busy, single-threaded conversation, including when users switch between earlier requests or interrupt an ongoing task.

Meta also said it has improved the model’s awareness of its own capabilities and limitations, helping it distinguish between what it knows and does not know and recognise when it encounters an obstacle rather than inventing an outcome.

Coding is another major area of improvement. Meta said Muse Spark 1.3 was trained on more long-horizon coding tasks and is better suited to engineering workflows.

Compared with Muse Spark 1.2, the new model is less verbose and requires fewer turns where they are unnecessary. In comparisons conducted by Meta engineers, Muse Spark 1.3 used around 20 per cent fewer tool calls and 25 per cent fewer tokens while completing coding tasks.

The company is also positioning the model for tasks that combine multiple tools and applications. A demonstration showed Muse Spark 1.3 handling a holiday campaign workflow involving Google Docs and a terminal.

Beyond performance, Meta has focused on safety for agentic use. Muse Spark 1.3 is claimed to have stronger resistance to adversarial inputs and prompt injections, while its handling of complex agentic tasks has been improved to better distinguish potentially irreversible actions.

The company said these changes are intended to give the model greater discretion when operating independently over longer periods.

Muse Spark 1.3 is currently available through Muse Code and Meta Model API. Meta also said its roadmap includes larger models and an open-weights release of Muse Spark.