$\"Run$ <\/p>\n

Muse Glimmer is gaining consideration within the native AI group and is being in contrast with Qwen’s 27B-class fashions. In lots of instances, it’s performing higher, particularly for native coding and agentic workflows.<\/p>\n

Meta seems robust within the open-model house, and with a couple of extra iterations, fashions like this might begin competing carefully with proprietary methods. As an AI fanatic, it is thrilling to have the ability to run this stage of AI domestically.<\/p>\n

On this information, we’ll run Muse Glimmer with llama.cpp<\/a><\/strong>, velocity it up with DFlash<\/strong>, and join it to Pi<\/strong> for native vibe coding. Will probably be capable of construct, check, and debug a mission straight from the terminal.<\/p>\n

1. Downloading Muse Glimmer<\/h2>\n
First, obtain the primary Muse Glimmer mannequin and its DFlash drafter from Hugging Face.<\/p>\n
Set up the Hugging Face CLI<\/a><\/strong>:<\/p>\n
curl -LsSf https:\/\/hf.co\/cli\/set up.sh | bash \necho 'export PATH=\"\/root\/.native\/bin:$PATH\"' >> ~\/.bashrc \nsupply ~\/.bashrc<\/code><\/pre>\nLog in:<\/p>\n
hf auth login<\/code><\/pre>\nCreate a mannequin listing:<\/p>\n
mkdir -p \/workspace\/muse-glimmer<\/code><\/pre>\nObtain the 16.8 GB primary mannequin<\/strong>:<\/p>\n
hf obtain meta-models\/Muse-Glimmer-30B-GGUF \n muse-glimmer-30B-kquant-17gb.gguf \n --local-dir \/workspace\/muse-glimmer<\/code><\/pre>\nObtain the 1.63 GB DFlash drafter<\/strong>:<\/p>\n
hf obtain meta-models\/Muse-Glimmer-30B-GGUF \n dflash-kquant.gguf \n --local-dir \/workspace\/muse-glimmer<\/code><\/pre>\nEach information will likely be saved in \/workspace\/muse-glimmer<\/code>.<\/p>\n
2. Putting in and Working llama.cpp<\/h2>\nSubsequent, set up llama.cpp with CUDA help and use it to serve Muse Glimmer with the DFlash drafter.<\/p>\n Set up and construct llama.cpp:<\/p>\n
cd \/workspace \n \ngit clone https:\/\/github.com\/ggml-org\/llama.cpp.git \ncd llama.cpp \ngit pull origin grasp \ncmake -B construct -DGGML_CUDA=ON -DCMAKE_BUILD_TYPE=Launch \ncmake --build construct --config Launch -j$(nproc) \nln -sf \"$(pwd)\/construct\/bin\/llama-server\" \/root\/.native\/bin\/llama-server<\/code><\/pre>\nTest the set up:<\/p>\n
git pull origin grasp \ncmake -B construct -DGGML_CUDA=ON -DCMAKE_BUILD_TYPE=Launch \ncmake --build construct --config Launch -j$(nproc) \nln -sf \"$(pwd)\/construct\/bin\/llama-server\" \/root\/.native\/bin\/llama-server<\/code><\/pre>\nIt’s best to see the put in llama.cpp model and construct info.<\/p>\n Now begin Muse Glimmer with DFlash speculative decoding<\/strong>:<\/p>\nllama-server \n -m \/workspace\/muse-glimmer\/muse-glimmer-30B-kquant-17gb.gguf \n -md \/workspace\/muse-glimmer\/dflash-kquant.gguf \n --spec-type draft-dflash \n --spec-draft-n-max 15 \n -ngl all \n --spec-draft-ngl all \n -fa on \n --ctx-size 16384 \n --alias muse \n --host 0.0.0.0 \n --port 8080 \n --jinja<\/code><\/pre>\n<\/p>\n Right here, llama.cpp masses the primary mannequin and the separate DFlash drafter onto the GPU, utilizing speculative decoding to enhance technology velocity.<\/p>\n 3. Testing Muse Glimmer<\/h2>\nAs soon as the server is operating, you’ll be able to check Muse Glimmer straight by means of the built-in llama.cpp Net UI<\/strong>.<\/p>\n Open:<\/p>\n http:\/\/localhost:8080\/<\/code><\/pre>\nIn my preliminary testing, I used to be getting round 46 tokens\/second<\/strong>, which is already fairly good.<\/p>\n <\/p>\n Throughout longer coding duties, I noticed speeds attain round 127 tokens\/second<\/strong>, making the mannequin really feel a lot sooner for agentic coding workflows.<\/p>\n <\/p>\n The outcomes had been combined, although. Muse Glimmer created an HTML sport for me, nevertheless it did not work significantly nicely. For this sort of activity, I nonetheless discovered Qwen3.8-27B<\/strong> noticeably higher at producing working HTML apps and video games.<\/p>\n <\/p>\n 4. Putting in Pi Coding Agent<\/h2>\nSubsequent, set up Pi<\/a><\/strong> and join it to the llama.cpp server operating Muse Glimmer.<\/p>\n Set up Pi:<\/p>\n curl -fsSL https:\/\/pi.dev\/set up.sh | sh<\/code><\/pre>\nThen set up Hugging Face’s llama.cpp extension:<\/p>\npi set up git:github.com\/huggingface\/pi-llama<\/code><\/pre>\nRestart your terminal after set up.<\/p>\n The pi-llama<\/code> extension robotically connects to:<\/p>\nhttp:\/\/localhost:8080\/v1<\/code><\/pre>\nIt detects the fashions being served by llama.cpp, so you don’t want to configure fashions.json<\/code> manually.<\/p>\n5. Beginning Native Vibe Coding<\/h2>\nNow create a mission and choose Muse Glimmer because the mannequin inside Pi.<\/p>\n Create an empty mission:<\/p>\nmkdir -p \/workspace\/glimmer-test \ncd \/workspace\/glimmer-test<\/code><\/pre>\nLaunch Pi:<\/p>\npi<\/code><\/pre>\nInside Pi, run:<\/p>\n\/mannequin<\/code><\/pre>\nSeek for:<\/p>\nllama-cpp<\/code><\/pre>\nThen choose:<\/p>\nmuse<\/code><\/pre>\n<\/p>\n Muse Glimmer ought to now be out there by means of Pi’s llama-cpp<\/code> supplier.<\/p>\n <\/p>\n 6. Testing Muse Glimmer as a Coding Agent<\/h2>\nLastly, give Muse Glimmer an entire coding activity and let it construct, check, and debug the mission itself.<\/p>\n I used this immediate:<\/p>\n \u00a0<\/p>\n \n\nConstruct an entire Python activity administration API from scratch utilizing FastAPI.<\/p>\n Necessities: – Create a clear mission construction. – Add endpoints to create, record, replace, and delete duties. – Use SQLite for persistence. – Add enter validation and error dealing with. – Add pytest assessments for all endpoints. – Create necessities.txt and README.md. – Run the assessments your self. – Repair any errors and rerun the assessments till all the pieces passes.<\/p>\n Don’t ask me to create information or run instructions for you. Construct and check the entire mission your self.\n<\/p>\n<\/blockquote>\n \u00a0<\/p>\n <\/p>\n Muse Glimmer constructed the mission in round 2 minutes<\/strong>.<\/p>\n To check it domestically:<\/p>\n pip set up -r necessities.txt \nuvicorn app.primary:app --reload<\/code><\/pre>\n<\/p>\n Open the API documentation at:<\/p>\n http:\/\/localhost:8000\/docs<\/code><\/pre>\n<\/p>\n As an alternative of manually testing each endpoint, I additionally requested Muse Glimmer to check the entire API itself and provides me a closing report.<\/p>\n <\/p>\n For native agentic coding, that is the place Muse Glimmer impressed me most. It was quick, dealt with multi-step duties nicely, and took only some seconds to establish and repair points throughout debugging.<\/p>\n Remaining Ideas<\/h2>\nMuse Glimmer is an effective indicator of how far native AI coding has come, particularly when Meta gives the official mannequin information and really helpful configuration. For me, it was very simple to arrange and begin utilizing.<\/p>\n There are nonetheless a couple of tough edges, however as Muse Glimmer, llama.cpp, DFlash, and the encircling tooling matures, I anticipate higher outcomes, sooner speeds, and stronger agentic coding efficiency domestically.<\/p>\n In case you have an RTX 3090, 4090, or 5090<\/strong>, I might extremely advocate making an attempt both Muse Glimmer or Qwen3.8 domestically<\/strong>. At this level, it’s turning into tougher to justify paying for each AI coding request or sharing your code and information with third-party companies.<\/p>\n Native fashions are already getting surprisingly near the expertise of fashions like GLM-5.2<\/strong>, and I feel the subsequent few iterations will make native AI coding much more compelling.<\/p>\n \u00a0 \u00a0<\/p>\n Abid Ali Awan<\/a><\/strong><\/b><\/a> (@1abidaliawan<\/a>) is a licensed information scientist skilled who loves constructing machine studying fashions. At present, he’s specializing in content material creation and writing technical blogs on machine studying and information science applied sciences. Abid holds a Grasp’s diploma in expertise administration and a bachelor’s diploma in telecommunication engineering. His imaginative and prescient is to construct an AI product utilizing a graph neural community for college kids battling psychological sickness.<\/p>\n<\/p><\/div>\n\n","protected":false},"excerpt":{"rendered":" Muse Glimmer is gaining consideration within the native AI group and is being in contrast with Qwen’s 27B-class fashions. In lots of instances, it’s performing higher, particularly for native coding and agentic workflows. Meta seems robust within the open-model house, and with a couple of extra iterations, fashions like this might begin competing carefully with […]<\/p>\n","protected":false},"author":2,"featured_media":18018,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[55],"tags":[1256,10286,10284,10285,1520,10283,733,1738],"class_list":["post-18016","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-machine-learning","tag-coding","tag-dflash","tag-glimmer","tag-llama-cpp","tag-local","tag-muse","tag-run","tag-vibe"],"_links":{"self":[{"href":"https:\/\/techtrendfeed.com\/index.php?rest_route=\/wp\/v2\/posts\/18016","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/techtrendfeed.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/techtrendfeed.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/techtrendfeed.com\/index.php?rest_route=\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/techtrendfeed.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=18016"}],"version-history":[{"count":1,"href":"https:\/\/techtrendfeed.com\/index.php?rest_route=\/wp\/v2\/posts\/18016\/revisions"}],"predecessor-version":[{"id":18017,"href":"https:\/\/techtrendfeed.com\/index.php?rest_route=\/wp\/v2\/posts\/18016\/revisions\/18017"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/techtrendfeed.com\/index.php?rest_route=\/wp\/v2\/media\/18018"}],"wp:attachment":[{"href":"https:\/\/techtrendfeed.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=18016"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/techtrendfeed.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=18016"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/techtrendfeed.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=18016"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}