Meta ships Muse Spark 1.3 with leaner tool use and longer-horizon agent runs
- Source
- x.com
- Date
We’re excited to release Muse Spark 1.3 with improved performance on agentic and coding tasks, and a focus on real-world usability. Key capabilities: → Sustains longer-horizon work across multiple workflows in a single thread → More actively collaborates with users: it asks clarifying questions, flags when it's stuck, confirms before consequential actions → Better calibrated on its own limits instead of hallucinating outcomes → ~20% fewer tool calls and ~25% fewer tokens vs. Muse Spark 1.2 in internal comparisons
Muse Spark 1.3 is rolling out today in Muse Code and Meta Model API at https://t.co/EMi0nXEvtz, with max reasoning coming soon after we finish safety testing. Learn more: https://t.co/I0IjQwpd8Q
Stay tuned for more updates soon, including bigger models, Muse Spark open weights, and more. 🚀
Context
AI at Meta said Muse Spark 1.3 can sustain longer-horizon work across multiple workflows within a single conversation thread, and is designed to collaborate more actively: asking clarifying questions, flagging when it gets stuck, and confirming before taking consequential actions. Meta also said the model is better about its own limits rather than guessing at outcomes it has not verified.
Meta reported roughly 20% fewer tool calls and 25% fewer per task than Muse Spark 1.2 in its own internal comparisons, figures the company has not published as an independent . The model is rolling out now in Muse Code and the Meta Model API, with a higher-reasoning mode to follow once Meta finishes safety testing.
- calibration — How well a model's confidence matches reality — a calibrated model saying "90% sure" is right about 90% of the time.
- token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.
- benchmark — A standard public test set for comparing AI models — the shared scoreboards behind every "model X beats model Y" claim.
- open weights — A model whose trained parameters are published for anyone to download and run — unlike API-only models you can access but never possess.
Checking sign-in…
Loading comments…