{ "archive_path": "archive/1792102052.88541", "base_url": "research.meta.ai/blog/muse-spark-open", "basename": "muse-spark-open", "bookmarked_date": "2026-10-15 22:07", "canonical": { "archive_org_path": "https://web.archive.org/web/research.meta.ai/blog/muse-spark-open", "dom_path": "output.html", "favicon_path": "favicon.ico", "git_path": "git/", "google_favicon_path": "https://www.google.com/s2/favicons?domain=research.meta.ai", "headers_path": "headers.json", "htmltotext_path": "htmltotext.txt", "index_path": "index.html", "media_path": "media/", "mercury_path": "mercury/content.html", "pdf_path": "output.pdf", "readability_path": "readability/content.html", "screenshot_path": "screenshot.png", "singlefile_path": "singlefile.html", "warc_path": "warc/", "wget_path": "research.meta.ai/blog/muse-spark-open.html" }, "domain": "research.meta.ai", "extension": "", "hash": "1MMV7PGF37Z8N0DWAXGV", "history": { "archive_org": [], "dom": [], "favicon": [ { "cmd": [ "curl", "--silent", "--location", "--compressed", "--max-time", "120", "--output", "favicon.ico", "--user-agent", "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/118.0.0.0 Safari/537.36 ArchiveBox/0.7.4 (+https://github.com/ArchiveBox/ArchiveBox/) curl/curl 8.14.1 (x86_64-pc-linux-gnu)", "https://www.google.com/s2/favicons?domain=research.meta.ai" ], "cmd_version": "curl 8.14.1 (x86_64-pc-linux-gnu)", "end_ts": "2026-10-15T22:12:48.946105+00:00", "index_texts": null, "output": "favicon.ico", "pwd": "/data/archive/1792102052.88541", "schema": "ArchiveResult", "start_ts": "2026-10-15T22:12:48.398389+00:00", "status": "succeeded" }, { "cmd": [ "curl", "--silent", "--location", "--compressed", "--max-time", "120", "--output", "favicon.ico", "--user-agent", "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/118.0.0.0 Safari/537.36 ArchiveBox/0.7.4 (+https://github.com/ArchiveBox/ArchiveBox/) curl/curl 8.14.1 (x86_64-pc-linux-gnu)", "https://www.google.com/s2/favicons?domain=research.meta.ai" ], "cmd_version": "curl 8.14.1 (x86_64-pc-linux-gnu)", "end_ts": "2026-10-16T01:41:29.370387+00:00", "index_texts": null, "output": "favicon.ico", "pwd": "/data/archive/1792102052.88541", "schema": "ArchiveResult", "start_ts": "2026-10-16T01:41:28.850515+00:00", "status": "succeeded" } ], "git": [], "headers": [ { "cmd": [ "curl", "--silent", "--location", "--compressed", "--head", "--max-time", "120", "--user-agent", "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/118.0.0.0 Safari/537.36 ArchiveBox/0.7.4 (+https://github.com/ArchiveBox/ArchiveBox/) curl/curl 8.14.1 (x86_64-pc-linux-gnu)", "https://research.meta.ai/blog/muse-spark-open" ], "cmd_version": "curl 8.14.1 (x86_64-pc-linux-gnu)", "end_ts": "2026-10-15T22:12:49.417059+00:00", "index_texts": null, "output": "headers.json", "pwd": "/data/archive/1792102052.88541", "schema": "ArchiveResult", "start_ts": "2026-10-15T22:12:48.963546+00:00", "status": "succeeded" }, { "cmd": [ "curl", "--silent", "--location", "--compressed", "--head", "--max-time", "120", "--user-agent", "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/118.0.0.0 Safari/537.36 ArchiveBox/0.7.4 (+https://github.com/ArchiveBox/ArchiveBox/) curl/curl 8.14.1 (x86_64-pc-linux-gnu)", "https://research.meta.ai/blog/muse-spark-open" ], "cmd_version": "curl 8.14.1 (x86_64-pc-linux-gnu)", "end_ts": "2026-10-16T01:41:29.958686+00:00", "index_texts": null, "output": "headers.json", "pwd": "/data/archive/1792102052.88541", "schema": "ArchiveResult", "start_ts": "2026-10-16T01:41:29.377827+00:00", "status": "succeeded" } ], "htmltotext": [], "media": [], "mercury": [ { "cmd": [ "/app/node_modules/@postlight/parser/cli.js", "https://research.meta.ai/blog/muse-spark-open" ], "cmd_version": "1.0.0", "end_ts": "2026-10-15T22:12:56.951586+00:00", "index_texts": null, "output": "mercury", "pwd": "/data/archive/1792102052.88541", "schema": "ArchiveResult", "start_ts": "2026-10-15T22:12:52.721949+00:00", "status": "succeeded" }, { "cmd": [ "/app/node_modules/@postlight/parser/cli.js", "https://research.meta.ai/blog/muse-spark-open" ], "cmd_version": "1.0.0", "end_ts": "2026-10-16T01:41:36.629294+00:00", "index_texts": null, "output": "mercury", "pwd": "/data/archive/1792102052.88541", "schema": "ArchiveResult", "start_ts": "2026-10-16T01:41:32.752285+00:00", "status": "succeeded" } ], "pdf": [], "readability": [ { "cmd": [ "/app/node_modules/readability-extractor/readability-extractor", "/tmp/tmpycqfce_c", "https://research.meta.ai/blog/muse-spark-open" ], "cmd_version": "0.0.11", "end_ts": "2026-10-15T22:12:52.711056+00:00", "index_texts": [ "Muse Spark Open is available today and brings leading agentic capabilities to your desktop for the first time on Apple M5 Ultra Mac Studios.\nToday, we\u2019re releasing the weights for Muse Spark Open, the most capable US open-weight model to date. Muse Spark Open is a natively multimodal model with 816B total parameters (42B active) and a 1M-token context window. Built for coding and agentic work, it excels at long-horizon tasks, complex software engineering, and multi-step reasoning.\nWe believe the future is for everyone, and that includes the freedom to run advanced models wherever, including at home. Muse Spark Open is the first model in its class to run on your desktop using two Apple M5 Ultra Mac Studios with the latest MLX framework.\nMuse Spark Open is available today through API platforms at Baseten, Fireworks\u00a0AI, Groq, Modal, Nebius, OpenRouter and Together\u00a0AI. Model weights are available for download on Amazon SageMaker AI, Google Cloud Model Garden, Hugging\u00a0Face, and Microsoft Foundry Models.\nIt runs on leading inference engines like SGLang, vLLM, TensorRT-LLM, Hugging\u00a0Face Transformers, Apple MLX, llama.cpp, and Unsloth with optimizations for AMD, Apple, Intel, and NVIDIA silicon. Developers can easily further customize the model using the PyTorch framework.Key Benchmarks\n\nFor more details about our evaluations, see the evaluation methodology report.\nOptimized for Apple Silicon\nLeveraging the latest version of MLX, Muse Spark Open was optimized to run on Mac\u00a0Studio with M5 Ultra, and together they\u2019re the best way to run a powerful model on your own desktop.\nApple\u2019s unified memory architecture lets the M5 Ultra's integrated GPU use up to 512 GB per Mac\u00a0Studio, so two machines can hold an entire model. And the chip\u2019s 1.2 TB/s of memory bandwidth is fast enough to read the small slice each token needs. The result is two Mac Studios behave like one supercomputer with a terabyte of unified memory.\nWe tested a range of agentic workflows running entirely on-device, using Muse Code powered by Muse Spark Open, including shipping real iOS changes to Ice Cubes (an open-source Mastodon client) \u2014 bug fixes, new features, full build \u2014 and driving DaVinci Resolve end to end, with the model reviewing footage frame by frame to find the strongest moments and crop them vertically.\nAvailable Now\nWe\u2019re sharing this work because we want everyone to have access to the best models. The open-weight version of Muse Spark ships with custom, auto-tunable Helion GPU kernels in SGLang, delivering 1.03-3.95\u00d7 kernel-level speedups on NVIDIA and AMD GPUs, and a 3% end-to-end speedup on AMD MI350.\nSafety and Preparedness\nReleasing open-weight models is part of how we bring the benefits of personal superintelligence to everyone, and part of how safety work advances across the field.\nOpen weights also carry their own considerations, and Muse Spark Open has additional safety post-training for that reason. The results demonstrate strong safeguards across risk categories measured.\nWe evaluated the model in line with our Meta Superintelligence Scaling Framework, covering the most severe risks, and it met the thresholds our framework requires for release. Detailed results are in the Muse Spark Open model card, with a full evaluation report available soon.\nLicense\nWe\u2019re releasing the weights under the Muse\u00a0Model\u00a0License with limited commercial terms, which allow many developers to use the model for free.\nGet Started\nYou can customize Muse Spark Open for your use case by leveraging PyTorch\u2019s TorchTitan to further tune the model. To run it on Apple\u00a0silicon, use MLX, which pulls the weights directly from Hugging\u00a0Face.\nMuse Spark Open is available now through Hugging\u00a0Face: https://huggingface.co/collections/meta-models/muse-spark-open\nMLX Adapter for running Muse Spark Open on Mac\u00a0Studio.\nHosted Inference Partners: Together\u00a0AI, Fireworks, Baseten, Modal, Groq, Nebius, and OpenRouter.\nBuilt or published something with Muse Spark Open? Share with us on the Hugging\u00a0Face Forums \u2014 we\u2019d love to see what you create." ], "output": "readability", "pwd": "/data/archive/1792102052.88541", "schema": "ArchiveResult", "start_ts": "2026-10-15T22:12:50.642523+00:00", "status": "succeeded" }, { "cmd": [ "/app/node_modules/readability-extractor/readability-extractor", "/tmp/tmp5ex_b_47", "https://research.meta.ai/blog/muse-spark-open" ], "cmd_version": "0.0.11", "end_ts": "2026-10-16T01:41:32.737686+00:00", "index_texts": [ "Muse Spark Open is available today and brings leading agentic capabilities to your desktop for the first time on Apple M5 Ultra Mac Studios.\nToday, we\u2019re releasing the weights for Muse Spark Open, the most capable US open-weight model to date. Muse Spark Open is a natively multimodal model with 816B total parameters (42B active) and a 1M-token context window. Built for coding and agentic work, it excels at long-horizon tasks, complex software engineering, and multi-step reasoning.\nWe believe the future is for everyone, and that includes the freedom to run advanced models wherever, including at home. Muse Spark Open is the first model in its class to run on your desktop using two Apple M5 Ultra Mac Studios with the latest MLX framework.\nMuse Spark Open is available today through API platforms at Baseten, Fireworks\u00a0AI, Groq, Modal, Nebius, OpenRouter and Together\u00a0AI. Model weights are available for download on Amazon SageMaker AI, Google Cloud Model Garden, Hugging\u00a0Face, and Microsoft Foundry Models.\nIt runs on leading inference engines like SGLang, vLLM, TensorRT-LLM, Hugging\u00a0Face Transformers, Apple MLX, llama.cpp, and Unsloth with optimizations for AMD, Apple, Intel, and NVIDIA silicon. Developers can easily further customize the model using the PyTorch framework.Key Benchmarks\n\nFor more details about our evaluations, see the evaluation methodology report.\nOptimized for Apple Silicon\nLeveraging the latest version of MLX, Muse Spark Open was optimized to run on Mac\u00a0Studio with M5 Ultra, and together they\u2019re the best way to run a powerful model on your own desktop.\nApple\u2019s unified memory architecture lets the M5 Ultra's integrated GPU use up to 512 GB per Mac\u00a0Studio, so two machines can hold an entire model. And the chip\u2019s 1.2 TB/s of memory bandwidth is fast enough to read the small slice each token needs. The result is two Mac Studios behave like one supercomputer with a terabyte of unified memory.\nWe tested a range of agentic workflows running entirely on-device, using Muse Code powered by Muse Spark Open, including shipping real iOS changes to Ice Cubes (an open-source Mastodon client) \u2014 bug fixes, new features, full build \u2014 and driving DaVinci Resolve end to end, with the model reviewing footage frame by frame to find the strongest moments and crop them vertically.\nAvailable Now\nWe\u2019re sharing this work because we want everyone to have access to the best models. The open-weight version of Muse Spark ships with custom, auto-tunable Helion GPU kernels in SGLang, delivering 1.03-3.95\u00d7 kernel-level speedups on NVIDIA and AMD GPUs, and a 3% end-to-end speedup on AMD MI350.\nSafety and Preparedness\nReleasing open-weight models is part of how we bring the benefits of personal superintelligence to everyone, and part of how safety work advances across the field.\nOpen weights also carry their own considerations, and Muse Spark Open has additional safety post-training for that reason. The results demonstrate strong safeguards across risk categories measured.\nWe evaluated the model in line with our Meta Superintelligence Scaling Framework, covering the most severe risks, and it met the thresholds our framework requires for release. Detailed results are in the Muse Spark Open model card, with a full evaluation report available soon.\nLicense\nWe\u2019re releasing the weights under the Muse\u00a0Model\u00a0License with limited commercial terms, which allow many developers to use the model for free.\nGet Started\nYou can customize Muse Spark Open for your use case by leveraging PyTorch\u2019s TorchTitan to further tune the model. To run it on Apple\u00a0silicon, use MLX, which pulls the weights directly from Hugging\u00a0Face.\nMuse Spark Open is available now through Hugging\u00a0Face: https://huggingface.co/collections/meta-models/muse-spark-open\nMLX Adapter for running Muse Spark Open on Mac\u00a0Studio.\nHosted Inference Partners: Together\u00a0AI, Fireworks, Baseten, Modal, Groq, Nebius, and OpenRouter.\nBuilt or published something with Muse Spark Open? Share with us on the Hugging\u00a0Face Forums \u2014 we\u2019d love to see what you create." ], "output": "readability", "pwd": "/data/archive/1792102052.88541", "schema": "ArchiveResult", "start_ts": "2026-10-16T01:41:31.030862+00:00", "status": "succeeded" } ], "screenshot": [], "singlefile": [], "title": [ { "cmd": [ "curl", "--silent", "--location", "--compressed", "--max-time", "120", "--user-agent", "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/118.0.0.0 Safari/537.36 ArchiveBox/0.7.4 (+https://github.com/ArchiveBox/ArchiveBox/) curl/curl 8.14.1 (x86_64-pc-linux-gnu)", "https://research.meta.ai/blog/muse-spark-open" ], "cmd_version": "curl 8.14.1 (x86_64-pc-linux-gnu)", "end_ts": "2026-10-15T22:12:50.634577+00:00", "index_texts": null, "output": "Introducing Muse Spark Open, The Most Capable US Open\u2011Weight Model to Date | Meta AI Research", "pwd": "/data/archive/1792102052.88541", "schema": "ArchiveResult", "start_ts": "2026-10-15T22:12:50.616248+00:00", "status": "succeeded" }, { "cmd": [ "curl", "--silent", "--location", "--compressed", "--max-time", "120", "--user-agent", "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/118.0.0.0 Safari/537.36 ArchiveBox/0.7.4 (+https://github.com/ArchiveBox/ArchiveBox/) curl/curl 8.14.1 (x86_64-pc-linux-gnu)", "https://research.meta.ai/blog/muse-spark-open" ], "cmd_version": "curl 8.14.1 (x86_64-pc-linux-gnu)", "end_ts": "2026-10-16T01:41:31.023181+00:00", "index_texts": null, "output": "Introducing Muse Spark Open, The Most Capable US Open\u2011Weight Model to Date | Meta AI Research", "pwd": "/data/archive/1792102052.88541", "schema": "ArchiveResult", "start_ts": "2026-10-16T01:41:31.005267+00:00", "status": "succeeded" } ], "wget": [ { "cmd": [ "wget", "--no-verbose", "--adjust-extension", "--convert-links", "--force-directories", "--span-hosts", "--no-parent", "-e", "robots=off", "--reject=js,mjs,woff,woff2,ttf,otf,eot,mp4,m4v,mov,webm,ogv,mkv,avi,flv", "--timeout=120", "--restrict-file-names=windows", "--page-requisites", "--user-agent=Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/118.0.0.0 Safari/537.36 ArchiveBox/0.7.4 (+https://github.com/ArchiveBox/ArchiveBox/) wget/GNU Wget 1.21.3", "--compression=auto", "--timestamping", "https://research.meta.ai/blog/muse-spark-open" ], "cmd_version": "GNU Wget 1.21.3", "end_ts": "2026-10-15T22:12:50.606974+00:00", "index_texts": null, "output": "research.meta.ai/blog/muse-spark-open.html", "pwd": "/data/archive/1792102052.88541", "schema": "ArchiveResult", "start_ts": "2026-10-15T22:12:49.433440+00:00", "status": "succeeded" }, { "cmd": [ "wget", "--no-verbose", "--adjust-extension", "--convert-links", "--force-directories", "--span-hosts", "--no-parent", "-e", "robots=off", "--reject=js,mjs,woff,woff2,ttf,otf,eot,mp4,m4v,mov,webm,ogv,mkv,avi,flv", "--timeout=120", "--restrict-file-names=windows", "--page-requisites", "--user-agent=Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/118.0.0.0 Safari/537.36 ArchiveBox/0.7.4 (+https://github.com/ArchiveBox/ArchiveBox/) wget/GNU Wget 1.21.3", "--compression=auto", "--timestamping", "https://research.meta.ai/blog/muse-spark-open" ], "cmd_version": "GNU Wget 1.21.3", "end_ts": "2026-10-16T01:41:30.995757+00:00", "index_texts": null, "output": "research.meta.ai/blog/muse-spark-open.html", "pwd": "/data/archive/1792102052.88541", "schema": "ArchiveResult", "start_ts": "2026-10-16T01:41:29.967953+00:00", "status": "succeeded" } ] }, "icons": null, "is_archived": true, "is_static": false, "latest": { "archive_org": null, "dom": null, "favicon": "favicon.ico", "git": null, "media": null, "pdf": null, "screenshot": null, "singlefile": null, "title": "Introducing Muse Spark Open, The Most Capable US Open\u2011Weight Model to Date | Meta AI Research", "warc": null, "wget": "research.meta.ai/blog/muse-spark-open.html" }, "link_dir": "/data/archive/1792102052.88541", "newest_archive_date": "2026-10-16T01:41:32.752285+00:00", "num_failures": 0, "num_outputs": 12, "oldest_archive_date": "2026-10-15T22:12:48.398389+00:00", "path": "/blog/muse-spark-open", "schema": "Link", "scheme": "https", "snapshot_id": "0498dddd-345f-424d-bda7-553fdf1b3c50", "sources": [ "/data/sources/1792102052-import.txt" ], "tags": ",war-wiki", "tags_str": "war-wiki", "timestamp": "1792102052.88541", "title": "Introducing Muse Spark Open, The Most Capable US Open\u2011Weight Model to Date | Meta AI Research", "updated": "2026-10-16T01:41:28.850372+00:00", "updated_date": "2026-10-16 01:41", "url": "https://research.meta.ai/blog/muse-spark-open" }