--- title: Fast, On Device Agentic AI with Muse Glimmer on... description: Meta has introduced Muse Glimmer, an open-weight 30B-parameter model distilled from Muse Spark, designed for on-device agentic workflows. ExecuTorch now... canonical: https://daily.dev/posts/fast-on-device-agentic-ai-with-muse-glimmer-on-executorch-pytorch-vg7lkn6ay twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Fast, On Device Agentic AI with Muse Glimmer on ExecuTorch – PyTorch | daily.dev og:description: Meta has introduced Muse Glimmer, an open-weight 30B-parameter model distilled from Muse Spark, designed for on-device agentic workflows. ExecuTorch now... og:url: https://daily.dev/posts/fast-on-device-agentic-ai-with-muse-glimmer-on-executorch-pytorch-vg7lkn6ay og:image: https://api.daily.dev/og/posts/Vg7lkN6AY.png og:image:alt: Fast, On Device Agentic AI with Muse Glimmer on ExecuTorch – PyTorch 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. # Fast, On Device Agentic AI with Muse Glimmer on ExecuTorch – PyTorch **[PyTorch](https://daily.dev/sources/pytorch)** · 5 min read · 0 upvotes · 0 comments ## Summary Meta has introduced Muse Glimmer, an open-weight 30B-parameter model distilled from Muse Spark, designed for on-device agentic workflows. ExecuTorch now provides end-to-end support for running it on NVIDIA GPUs and Apple Silicon Macs. Key features include DFlash parallel diffusion-based speculative decoding (achieving 52.8% throughput improvement on M5 Pro), native GGUF/K-quant execution, 128K+ token context via a hybrid global/sliding-window KV-cache (only 13 of 52 layers are global), and multimodal text+image inputs. Prebuilt PTE artifacts are available on Hugging Face. The post details backend-specific optimizations for CUDA (CUDA graphs, packed K-quant kernels, FlashDecoding++) and MLX/Metal (RMSNorm, RoPE, SDPA lowered to native ops), plus an agentic harness supporting multi-session isolation, XML tool-call parsing, and Harmony chat templating. Video input and cross-session prefix sharing are planned for future releases. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What performance improvement does DFlash speculative decoding give for Muse Glimmer on Apple Silicon? DFlash speculative decoding achieves a 52.8% throughput improvement over solo (greedy) decoding for Muse Glimmer on an M5 Pro with 64 GiB RAM. Solo decoding reaches 21.6 tokens/second for text-image input, while DFlash reaches 33.0 tokens/second, with no quality regression reported. _Engineers optimizing on-device LLM inference track speculative decoding benchmarks like these on daily.dev._ ### How does Muse Glimmer handle 128K token context efficiently on edge devices? Muse Glimmer uses a hybrid KV-cache architecture: only 13 of its 52 layers are global-attention layers, while the remaining 39 use sliding-window attention. This limits KV-cache growth to the global layers, making 128K+ token contexts practical on edge hardware. ExecuTorch supports this architecture natively for both CUDA and MLX backends. _Developers shipping long-context models to edge hardware find architecture trade-offs like this covered on daily.dev._ ### How does ExecuTorch export Muse Glimmer GGUF checkpoints for CUDA and Apple Silicon? ExecuTorch exports directly from released GGUF checkpoints via its torch.export-based ahead-of-time stack. For CUDA, it compiles and autotunes Triton kernels for the detected GPU architecture, mapping Q4_K/Q5_K/Q6_K to packed INT4/5/6 with dp4a GEMV kernels. For MLX, it uses repacked or fused Metal kernels, and merges adjacent sub-blocks with identical scale/min into larger group sizes up to 128 when lossless. _Teams deploying quantized models to multiple backends follow ExecuTorch developments on daily.dev._ ## Similar posts on daily.dev - [Fine-tune Meta AI's Muse Glimmer 100% locally](https://daily.dev/posts/fine-tune-meta-ai-s-muse-glimmer-100-locally-nvrlxatlc) · Daily Dose of Data Science \| Avi Chawla \| Substack · 0 upvotes · 0 comments - [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), [#pytorch](https://daily.dev/tags/pytorch), [#local-ai](https://daily.dev/tags/local-ai), [#ai-inference](https://daily.dev/tags/ai-inference) [View this post on daily.dev](https://daily.dev/posts/fast-on-device-agentic-ai-with-muse-glimmer-on-executorch-pytorch-vg7lkn6ay) ```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":"Fast, On Device Agentic AI with Muse Glimmer on ExecuTorch – PyTorch","url":"https://daily.dev/posts/fast-on-device-agentic-ai-with-muse-glimmer-on-executorch-pytorch-vg7lkn6ay","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/fast-on-device-agentic-ai-with-muse-glimmer-on-executorch-pytorch-vg7lkn6ay"},"datePublished":"2026-08-10T13:46:52.551Z","dateModified":"2026-08-10T14:17:18.992Z","description":"Meta has introduced Muse Glimmer, an open-weight 30B-parameter model distilled from Muse Spark, designed for on-device agentic workflows. ExecuTorch now...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/6c6e8ee318693c4de6f5f85f45160b68?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/6c6e8ee318693c4de6f5f85f45160b68?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"PyTorch","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":"PyTorch","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/5fdab3d75a834a97a52d433d9c7e5ff9","url":"https://daily.dev/sources/pytorch"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/fast-on-device-agentic-ai-with-muse-glimmer-on-executorch-pytorch-vg7lkn6ay","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"llm,pytorch,local-ai,ai-inference","timeRequired":"PT5M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"PyTorch","item":"https://daily.dev/sources/pytorch"},{"@type":"ListItem","position":3,"name":"Fast, On Device Agentic AI with Muse Glimmer on ExecuTorch – PyTorch"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/fast-on-device-agentic-ai-with-muse-glimmer-on-executorch-pytorch-vg7lkn6ay#faq","mainEntity":[{"@type":"Question","name":"What performance improvement does DFlash speculative decoding give for Muse Glimmer on Apple Silicon?","acceptedAnswer":{"@type":"Answer","text":"DFlash speculative decoding achieves a 52.8% throughput improvement over solo (greedy) decoding for Muse Glimmer on an M5 Pro with 64 GiB RAM. Solo decoding reaches 21.6 tokens/second for text-image input, while DFlash reaches 33.0 tokens/second, with no quality regression reported. Engineers optimizing on-device LLM inference track speculative decoding benchmarks like these on daily.dev."}},{"@type":"Question","name":"How does Muse Glimmer handle 128K token context efficiently on edge devices?","acceptedAnswer":{"@type":"Answer","text":"Muse Glimmer uses a hybrid KV-cache architecture: only 13 of its 52 layers are global-attention layers, while the remaining 39 use sliding-window attention. This limits KV-cache growth to the global layers, making 128K+ token contexts practical on edge hardware. ExecuTorch supports this architecture natively for both CUDA and MLX backends. Developers shipping long-context models to edge hardware find architecture trade-offs like this covered on daily.dev."}},{"@type":"Question","name":"How does ExecuTorch export Muse Glimmer GGUF checkpoints for CUDA and Apple Silicon?","acceptedAnswer":{"@type":"Answer","text":"ExecuTorch exports directly from released GGUF checkpoints via its torch.export-based ahead-of-time stack. For CUDA, it compiles and autotunes Triton kernels for the detected GPU architecture, mapping Q4_K/Q5_K/Q6_K to packed INT4/5/6 with dp4a GEMV kernels. For MLX, it uses repacked or fused Metal kernels, and merges adjacent sub-blocks with identical scale/min into larger group sizes up to 128 when lossless. Teams deploying quantized models to multiple backends follow ExecuTorch developments on daily.dev."}}]} ```