AI at Meta

Today we're introducing Muse Ember, our smallest and most efficient model yet. Muse Ember is built for agentic workloads, like tool calls, coding subtasks, routing, extraction and long document parsing. It achieves performance comparable to Claude Haiku 5.5 at roughly one-third of the cost per task, while scoring higher than GPT-6 Luna across all of Luna's reasoning settings. Muse Ember is available today on the Meta Model API, at $0.10 per million input tokens and $0.50 per million output tokens. Learn more: https://lnkd.in/eneTTcE2

Today we're bringing specialized models trained using Muse Wonder to the Muse app and Meta AI, giving users access to some of Muse Wonder's frontier capabilities for complex tasks at work and in everyday life. These specialized models include Muse Wonder's Active Memory and Active Focus, helping them stay aligned with your goals as tasks unfold: → Active Memory helps recognize when earlier information matters and distinguish what's still relevant from what's become stale → Active Focus helps recover from setbacks or find alternatives when constraints change In Muse, the improved models show a 2.4x reduction in forgetting information critical to your goals and a 67% improvement in getting unstuck as constraints evolve. That means less repeating yourself and fewer dead ends when plans change. In Meta AI, these improvements help it break complex problems into steps, ask sharper follow-up questions and adapt when you change direction. Read more on our research blog: https://go.meta.me/4kCPcHW

During Connect, we previewed Muse Realtime Avatar, our real-time embodiment technology that turns Muse Realtime Voice into expressive, interactive avatars. Today, we’re rolling it out in the Muse app in the US, giving people a more natural way to engage with their personal agents. Learn more about Muse Realtime Avatar on our research blog: https://go.meta.me/47ZSNso

🛎️ We’ve published ODAC25 - the largest open-access dataset to date for carbon capture materials - in Nature Chemical Engineering, together with Meta FAIR, Georgia Institute of Technology and our academic partners. Direct Air Capture (DAC) relies on materials, including porous materials called Metal-Organic Frameworks (MOFs) to pull CO2 from the air. But simulating how they perform under real-world conditions has historically been a major computational bottleneck.  To model how materials behave under real-world conditions, AI models require millions of high-precision quantum chemistry calculations. Here’s how CuspAI contributed to the final dataset - which included 60M calculations across 15,000 MOFs: 1. CuspAI contributed new, synthetic MOF structures to the dataset, designed by one of its proprietary generative AI models. These MOFs extended the dataset towards more complex geometries - with larger pores, higher surface areas and lower densities than known experimental structures. 2. We released a simulation tool that can accurately simulate how gases interact with materials under complex conditions. Generative AI and large-scale simulation can rapidly expand the frontiers of materials science not just across carbon capture - but across industries. Find out more in the comments...

How can Muse models help you build and scale? Join us tomorrow, Thursday, Oct 29 at 9:30am PT to explore coding and agentic use cases with Muse Wonder and Muse Spark 1.3, learn how each model fits into your stack, and get direct feedback from Meta experts in live Q&A. RSVP: bit.ly/3U3z7R3

Today we're announcing support for university research at Caltech, The Ohio State University College of Medicine, Louisiana State University, Indiana University Bloomington, and UTHealth Houston to help advance breakthroughs across science and medicine. As part of the support for university researchers, we’ll provide access to our latest Muse models, including Muse Wonder and Muse Spark; open weight models like Muse Glimmer and Muse Spark Open; computer vision models such as SAM, V-JEPA, and DINO; as well as SeamlessM4T, our multilingual communication model. Here’s a look at the hard problems these five universities are tackling: Caltech: Expanding our partnership with Caltech to test whether large language models can accelerate innovative research in mathematics and physical sciences – pushing the boundaries of what AI can do in rigorous scientific domains The Ohio State University College of Medicine: Developing AI workflows to assist radiologists screening X-rays, building AI tools for earlier heart failure detection, and building clinician-guided AI tools for Alzheimer’s diagnosis. Louisiana State University: Implementing AI to improve non-contact infant tracheostomy monitoring, neurosurgical outcomes, VR training for medical professionals, and more.  Indiana University Bloomington: Applying open-weight models to improve maternal-newborn care and learning, with a focus on tools that work with limited connectivity and accommodate diverse cognitive needs. UTHealth Houston: Using Muse Glimmer, one of our smallest open-weight models, to support projects that improve treatment for brain cancer and agentic benchmarking in healthcare settings. All healthcare research is conducted independently by participating universities, ensuring patient and clinical data remain under their control. We’re excited to support these five universities as they use our models to accelerate the important work they’re doing. Read more on our research blog: https://go.meta.me/47G6wV3

Muse Wonder is now available on Cline, Cloudflare, Cognition, Cursor, OpenRouter, Oracle and Vercel - with more partners coming soon. We look forward to seeing what the community builds with Muse Wonder! Start building 👉 https://bit.ly/4zKErb3

We’re proud to share that Meta Omnilingual Automatic Speech Recognition (ASR) has been named one of TIME’s Best Inventions of 2026! Meta Omnilingual ASR is a suite of models providing ASR capabilities for over 1,600 languages, including 500 low-coverage languages never before served by any ASR system. While most ASR systems focus on a limited set of languages that are well-represented on the internet, this release marked a major step toward building a truly universal transcription system. Learn more about it here: https://lnkd.in/e5vhkqkb

Today we’re releasing Muse Spark Open, the most capable US open-weight model to date. Built for coding and agentic work, it excels at long-horizon tasks, complex software engineering, and multi-step reasoning. Key Model Specs: → 816B total / 42B active parameters → 1M token context window → Natively multimodal across text, image, video, audio To help make advanced models more accessible, Muse Spark Open is optimized to run on two Apple M5 Ultra Mac Studios with the latest MLX framework. It also runs on leading inference engines like SGLang, vLLM, TensorRT-LLM, Hugging Face Transformers, llama.cpp, and Unsloth with optimizations for AMD, Apple, Intel, and NVIDIA silicon. Developers can further customize the model with PyTorch, including TorchTitan for fine-tuning. We’re releasing the weights under the Muse Model License, with limited commercial terms that let many developers use it for free. Learn more on our research blog: https://lnkd.in/eUiykJ6Q Download the weights: https://lnkd.in/ef-SzyDs

We’re exploring how AI can accelerate health science research. While AI has made significant breakthroughs in formal sciences like mathematics, health data science depends on imperfect real-world data and consequential choices about cohorts, variables, and causal assumptions. To explore how AI can assist health researchers, we developed a human-led research setup powered by an internal coding scaffold using Muse Spark 1.3 and Muse Wonder. The system was given access to publicly available data as well as a licensed database of 5 million retrospective de-identified electronic health records. A cohort of medical researchers across cardiology, epidemiology, infectious disease, internal medicine, and oncology used the scaffold to conduct original  research over just four weeks. Highlights:  1️⃣ Rapid Execution: Researchers generated completed manuscript drafts in as little as 7 days from initial prompt.  2️⃣ Significant Productivity Gains: Experts unanimously reported substantial productivity improvements, reaching up to 10x in best-case scenarios compared to traditional workflows.  3️⃣ Human-in-the-Loop Governance: Domain experts remained fully responsible for study design, causal assumptions, and scientific conclusions, while AI accelerated hypothesis generation, code execution, and manuscript drafting. Today, we’re sharing 10 research manuscripts produced through this exploration, covering topics across infective endocarditis, breast cancer treatment sequences, bloodstream infections, emergency care trends, and more. We are also sharing a report written by these experts detailing their experience and learnings from doing research with Muse Spark and Muse Wonder. Read more in our research blog: https://go.meta.me/3U06wvD