--- title: Run Local Agentic AI Workflows with Meta’s Muse Glimmer... description: Meta has released Muse Glimmer, a 30B open-weight dense model with a 120K+ context window designed for local agentic AI workflows. Unlike chat-optimized LLMs,... canonical: https://daily.dev/posts/run-local-agentic-ai-workflows-with-meta-s-muse-glimmer-on-nvidia-eme2jejbx twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Run Local Agentic AI Workflows with Meta’s Muse Glimmer on NVIDIA | daily.dev og:description: Meta has released Muse Glimmer, a 30B open-weight dense model with a 120K+ context window designed for local agentic AI workflows. Unlike chat-optimized LLMs,... og:url: https://daily.dev/posts/run-local-agentic-ai-workflows-with-meta-s-muse-glimmer-on-nvidia-eme2jejbx og:image: https://api.daily.dev/og/posts/EmE2jEJbx.png og:image:alt: Run Local Agentic AI Workflows with Meta’s Muse Glimmer on NVIDIA og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Run Local Agentic AI Workflows with Meta’s Muse Glimmer on NVIDIA **[NVIDIA Developer](https://daily.dev/sources/nvidiadev)** · 4 min read · 1 upvotes · 0 comments ## Summary Meta has released Muse Glimmer, a 30B open-weight dense model with a 120K+ context window designed for local agentic AI workflows. Unlike chat-optimized LLMs, its dense architecture activates all parameters per token, delivering reliable instruction following, long-context coherence, and predictable latency suited for multi-step agent tasks. It achieves over 20K tokens/sec on a single GPU and fits within the VRAM of a single NVIDIA GPU without sharding or CPU offloading. Supported hardware includes the RTX 5090, DGX Spark, DGX Station, and Jetson for edge deployments. Deployment options include SGLang, vLLM, and NVIDIA NIM containers. Fine-tuning is supported via NeMo AutoModel with SFT and LoRA out of the box, plus reinforcement learning via NeMo RL. Weights are available on HuggingFace. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What is Meta's Muse Glimmer model and what is it optimized for? Muse Glimmer is a 30B-parameter open-weight dense model from Meta with a context window over 120K tokens, built specifically for local agentic AI workloads rather than chat-style conversations. It uses a dense architecture that activates every parameter per token, avoiding mixture-of-experts routing overhead, which gives it reliable instruction following, long-context coherence, and predictable latency for multi-step agent tasks like coding or documentation revision. _Developers picking a local agentic model can track releases like this one through daily.dev._ ### How fast does Muse Glimmer run on NVIDIA GPUs? Muse Glimmer delivers up to 20K tokens/sec on a single GPU and over 20 tokens/sec per GPU on NVIDIA Blackwell Ultra at BF16/NVF4 precision. A single Blackwell Ultra GPU can hold the full 30B model in VRAM with headroom for large KV cache buffers, supporting high concurrency and low latency for always-on agentic workloads without model sharding or CPU offloading. _Anyone benchmarking local inference throughput can follow model performance updates via daily.dev._ ### What NVIDIA hardware can run Meta's Muse Glimmer model locally? Muse Glimmer runs locally on NVIDIA GeForce RTX 5090 (32GB VRAM), DGX Spark, DGX Station, and Jetson edge devices. DGX Spark uses NVLink for high-speed memory access and NIM containers for one-command deployment, DGX Station brings rack-scale Blackwell Ultra compute for air-gapped enterprise setups, and Jetson extends inference to robotics and embedded systems requiring network isolation. _Teams weighing hardware for on-device AI agents can keep up with platform options through daily.dev._ ## Similar posts on daily.dev - [Muse Glimmer 30B runs on my 32GB GPU, handles its own tool calls, and never leaves the machine](https://daily.dev/posts/muse-glimmer-30b-runs-on-my-32gb-gpu-handles-its-own-tool-calls-and-never-leaves-the-machine-vihln3unw) · XDA Developers · 1 upvotes · 0 comments --- Tags: [#llm](https://daily.dev/tags/llm), [#ai-agents](https://daily.dev/tags/ai-agents), [#deep-learning](https://daily.dev/tags/deep-learning), [#nvidia](https://daily.dev/tags/nvidia) [View this post on daily.dev](https://daily.dev/posts/run-local-agentic-ai-workflows-with-meta-s-muse-glimmer-on-nvidia-eme2jejbx) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Run Local Agentic AI Workflows with Meta’s Muse Glimmer on NVIDIA","url":"https://daily.dev/posts/run-local-agentic-ai-workflows-with-meta-s-muse-glimmer-on-nvidia-eme2jejbx","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/run-local-agentic-ai-workflows-with-meta-s-muse-glimmer-on-nvidia-eme2jejbx"},"datePublished":"2026-08-10T13:29:15.741Z","dateModified":"2026-09-13T19:31:25.306Z","description":"Meta has released Muse Glimmer, a 30B open-weight dense model with a 120K+ context window designed for local agentic AI workflows. Unlike chat-optimized LLMs,...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/9c65a68f6905b87da50e3a831ef040b3?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/9c65a68f6905b87da50e3a831ef040b3?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"NVIDIA Developer","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"NVIDIA Developer","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/86e45aab42ba48ce83103d01b1119910","url":"https://daily.dev/sources/nvidiadev"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/run-local-agentic-ai-workflows-with-meta-s-muse-glimmer-on-nvidia-eme2jejbx","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"llm,ai-agents,deep-learning,nvidia","timeRequired":"PT4M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"NVIDIA Developer","item":"https://daily.dev/sources/nvidiadev"},{"@type":"ListItem","position":3,"name":"Run Local Agentic AI Workflows with Meta’s Muse Glimmer on NVIDIA"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/run-local-agentic-ai-workflows-with-meta-s-muse-glimmer-on-nvidia-eme2jejbx#faq","mainEntity":[{"@type":"Question","name":"What is Meta's Muse Glimmer model and what is it optimized for?","acceptedAnswer":{"@type":"Answer","text":"Muse Glimmer is a 30B-parameter open-weight dense model from Meta with a context window over 120K tokens, built specifically for local agentic AI workloads rather than chat-style conversations. It uses a dense architecture that activates every parameter per token, avoiding mixture-of-experts routing overhead, which gives it reliable instruction following, long-context coherence, and predictable latency for multi-step agent tasks like coding or documentation revision. Developers picking a local agentic model can track releases like this one through daily.dev."}},{"@type":"Question","name":"How fast does Muse Glimmer run on NVIDIA GPUs?","acceptedAnswer":{"@type":"Answer","text":"Muse Glimmer delivers up to 20K tokens/sec on a single GPU and over 20 tokens/sec per GPU on NVIDIA Blackwell Ultra at BF16/NVF4 precision. A single Blackwell Ultra GPU can hold the full 30B model in VRAM with headroom for large KV cache buffers, supporting high concurrency and low latency for always-on agentic workloads without model sharding or CPU offloading. Anyone benchmarking local inference throughput can follow model performance updates via daily.dev."}},{"@type":"Question","name":"What NVIDIA hardware can run Meta's Muse Glimmer model locally?","acceptedAnswer":{"@type":"Answer","text":"Muse Glimmer runs locally on NVIDIA GeForce RTX 5090 (32GB VRAM), DGX Spark, DGX Station, and Jetson edge devices. DGX Spark uses NVLink for high-speed memory access and NIM containers for one-command deployment, DGX Station brings rack-scale Blackwell Ultra compute for air-gapped enterprise setups, and Jetson extends inference to robotics and embedded systems requiring network isolation. Teams weighing hardware for on-device AI agents can keep up with platform options through daily.dev."}}]} ```