--- title: "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 source_url: https://pytorch.org/blog/fast-ondevice-agentic-ai-with-executorch type: article source: "PyTorch" published: 2026-08-10T13:46:52.551Z updated: 2026-08-10T14:17:18.992Z tags: ["llm", "pytorch", "local-ai", "ai-inference"] reading_time: 5 upvotes: 0 comments: 0 language: 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._ --- 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)