Meta releases 816B parameter Muse Spark Open model with custom license
ByPulseAugur Editorial·[18 sources]·
Meta has released Muse Spark Open, an 816-billion-parameter mixture-of-experts model, with approximately 42 billion parameters active per token for efficiency. This model, designed for coding and agentic tasks with a 1M-token context window, scores 39 on the Artificial Analysis Intelligence Index, positioning it as the top-scoring US-built open-weight model. However, its release under a custom Muse Model License, rather than standard open-source licenses like Apache 2.0 or MIT, includes attribution requirements and a $20 million revenue threshold for separate agreements with Meta.
AI
IMPACT
Accelerates development of AI agents and coding tools by providing a powerful, albeit restrictively licensed, open-weight model.
RANK_REASON
Frontier-lab model release with system card and custom license terms.
<h2>Liquid AI LFM3</h2> <p><a href="https://ollama.com/library/lfm3" rel="nofollow">LFM3</a> is Liquid AI's new family of small models built for local agents, tool calling, and structured output, on Liquid's new tri-hybrid architecture. The family includes:</p> <ul> <li><a href="…
arXiv:2610.32637v1 Announce Type: new Abstract: Fine-tuning the non-blank predictor of a factorized transducer on target-domain text improves recognition on that domain but degrades performance on the source domain. A single interpolation weight between base and adapted predictor…
arXiv:2610.30685v1 Announce Type: new Abstract: Large distribution shifts remain a primary obstacle for domain generalization and unsupervised adaptation. Although gradual domain adaptation alleviates this difficulty by introducing ordered intermediate domains, most existing appr…
arXiv cs.LG
TIER_1English(EN)·Zijie Liu, Zhen Tan, Charles Fleming, Tianlong Chen·
arXiv:2610.28892v1 Announce Type: new Abstract: Adapting language models to new domains during post-training risks degrading previously acquired capabilities, a phenomenon commonly known as catastrophic forgetting. Existing mitigation methods largely treat forgetting as a consequ…
Policy-Mask learns a budget-conditioned selection policy over diffusion timesteps, network blocks, and channels to extract compact, task-optimal features for discriminative vision tasks.
Hugging Face Trending Models
TIER_1(SO)·MiniMaxAI·
arXiv:2610.30680v1 Announce Type: new Abstract: Domain generalization (DG) aims to improve the generalization of the model to any unseen domains. Targeting this goal, existing DG methods mainly focus on learning domain-invariant features or prompts from the given source domains. …
arXiv cs.CV
TIER_1English(EN)·Ivan Listopadov, Maksim Nakhodnov, Aibek Alanov·
arXiv:2610.28889v1 Announce Type: new Abstract: Selecting intermediate features from a frozen diffusion model for discriminative tasks is non-trivial because the useful signal is distributed across the denoising trajectory, network depth, and channels. In practice, this requires …
arXiv cs.CV
TIER_1English(EN)·Ben Vardi, Dana Schonberger, Yuval Friedmann, Zohar Yakhini, Iris Barshack, Alexander Loebel, Ariel Shamir·
arXiv:2610.28007v1 Announce Type: new Abstract: Pathology foundation models (PFMs) support a wide range of histopathology tasks on whole-slide images through multiple instance learning (MIL). Despite recent progress, these models remain sensitive to cross-hospital distribution sh…
Medium — AI coding tag
TIER_1English(EN)·Andre Burgstahler·
<p>Meta just did something it hasn't done in over two years: it released the weights of a frontier-class model. <strong>Muse Spark Open</strong>, launched October 15, 2026 by Meta Superintelligence Labs, is an 816-billion-parameter mixture-of-experts model — but only about 42 bil…
