Meta just dropped its most powerful AI model yet, and the company thinks it can finally go head-to-head with the biggest names in the AI race. Developers can access and pay for Muse Spark 1.3, the latest update to Meta’s flagship model. It won’t stay locked away for long either, as the update is set to roll out across Meta’s apps, including Instagram, Facebook, and Meta AI, in the coming weeks.
How good is Meta’s new AI model?
Meta’s Chief AI Officer Alexandr Wang called this update the company’s “biggest jump so far on model performance,” pointing to major gains in coding and agentic tasks, basically the AI’s ability to complete tasks on your behalf.
Wang told Bloomberg that Muse Spark 1.3 is competitive with Anthropic’s Claude Fable 5.1 and even beats OpenAI’s GPT-5.6 Sol at coding, though OpenAI’s upcoming Astra model looms on the horizon.
Of course, take these comparisons with a grain of salt. Benchmarks can be gamed, and a model that shines in one task might stumble in another.
What changed under the hood?
Wang broke down a few upgrades that make Muse Spark 1.3 stand out. It needs 25% fewer tokens to get the job done, meaning it runs more efficiently. It can juggle multiple workflows at once instead of forcing separate sessions, and it holds on to context better across long, complicated instructions. Meta also says the model now recognizes its own limitations and will pause to ask for clarification before taking any action that can’t be undone.
Meta hasn’t decided whether it will release the model’s weights, the blueprint that lets outside developers build on top of it, though the older Muse Spark 1.2 weights are still headed for release.
As for pricing, developers won’t pay more than they did for Muse Spark 1.2. Wang told Bloomberg that adoption on Meta’s developer platform has been strong, with some users burning through trillions of tokens every week.
Meta is still deep in its AI spending spree, but with Muse Spark 1.3, the company is signaling it’s closing the gap, one update at a time.
