816B total parameters and 42B active, with text, image, audio and video inputs and text output. Self-host under the Muse Model License; published quantized examples still occupy hundreds of GB of memory. AA at xhigh scores 39, separately from Spark 1.3 and Wonder.
Each model uses its highest tested setting, with fallback conditions retained. Original K3 remains at max44 from its readable individual page; the aggregate page now lists K3.1 Pro separately. Scores use one evaluator. The index mainly evaluates English text tasks; examine Chinese and image capabilities separately.
On October 15, Meta released Muse Spark Open for deployment, modification and fine-tuning under its license, continuing Llama’s route of putting models in other developers’ hands.
Independent text evaluation · xhigh · distinct from Spark 1.3
Coding, debugging and codebase understanding advance through co-training with Muse Code; the model and terminal agent launch together as distinct layers.
The new series emphasizes Active Memory, Active Focus, long-running tasks and coding. Wonder opens through Meta Model API and Muse Code; Full is for Muse Code subscribers. Personal Muse integration remains a later rollout.
The 816B-total, 42B-active multimodal model releases weights under the Muse Model License. It derives from August’s Spark with additional safety post-training, rather than the same weights as Spark 1.3 or Wonder.