{
  "@context": "https://schema.org",
  "@id": "https://aggr.aymericbeaumet.com/items/zml-ai/2026-10-22-zml-llmd-20261023-0/#webpage",
  "@type": [
    "WebPage",
    "ArchiveComponent"
  ],
  "articleBody": "[Back to posts](https://zml.ai/posts/)\n\nArticle\n\nMore speed, more models, more fun.\n\nOctober 22, 2026 inference / serving / NVIDIA / AMD / TPU / Metal / Intel\n\nWe are pleased to announce the release of `zml/llmd 20261023.0`.\n\nWhile last week’s release (`zml/llmd 20261015.0`) was a minor release with a focus on performance, this week’s release builds on the gains we made and brings more speed, more models, and more fun.\n\nAnd so this week, we bring new model series to zml/llmd.\n\n## Qwen/Qwen3.8-Flash-Next\n\nThis release brings the [Qwen/Qwen3.8-Flash-Next series](https://huggingface.co/collections/Qwen/qwen38-flash-next) model series to zml/llmd.\n\nWe have tested it on NVIDIA and AMD GPU, the only ones we got access to that could host the official release. Google is still telling us there are no TPUs available.\n\nDecode throughput, Qwen/Qwen3.8-Flash-Next, bs=16, zml/llmd 20261023.0 (tok/s)\n\n## XiaomiMiMo/MiMo-V2.6\n\nThis release brings the [XiaomiMiMo/MiMo-V2.6 series](https://huggingface.co/collections/XiaomiMiMo/mimo-v26) in both RL and MOPD flavour.\n\nDecode throughput, XiaomiMiMo/MiMo-V2.6-Flash-MOPD, bs=16, zml/llmd 20261023.0 (tok/s)\n\n## Faster DeepSeek-V4.1-Flash\n\nThe hill climb continues. We improved a number of fundamental building blocks and thus DeepSeek-V4.1-Flash got faster.\n\nToken Generation, deepseek-ai/DeepSeek-V4.1-Flash, bs=1, 4xGB300 (tok/s)\n\n## Try it today\n\n**NVIDIA**\n\n```bash\ndocker run -p 8000:8000 --shm-size=256GB --gpus=all -e HF_TOKEN -it zmlai/llmd:cuda \\\n    --model=hf://Qwen/Qwen3.8-Flash-Next\n```\n\n**AMD**\n\n```bash\ndocker run -p 8000:8000 --shm-size=256GB --device=/dev/kfd --device=/dev/dri -e HF_TOKEN -it zmlai/llmd:rocm \\\n    --model=hf://Qwen/Qwen3.8-Flash-Next\n```\n\n**Intel**\n\n```bash\ndocker run -p 8000:8000 --device=/dev/dri -e HF_TOKEN -it zmlai/llmd:oneapi \\\n    --model=hf://Qwen/Qwen3.8-27B\n```\n\n**Google TPU**\n\n```bash\ndocker run --net=host --privileged -e HF_TOKEN -it zmlai/llmd:tpu \\\n    --model=hf://Qwen/Qwen3.8-27B\n```\n\n**Apple Metal**\n\n```bash\nbrew install zml/zml/llmd\nllmd --model=hf://nvidia/Qwen3.8-27B-NVFP4 --batch-size=1\n```\n",
  "authors": [],
  "captured_at": "2026-10-23T18:40:08.262074691+00:00",
  "category": "ai",
  "content_html": "<p><a target=\"_blank\" rel=\"noopener noreferrer\" href=\"https://zml.ai/posts/\">Back to posts</a></p>\n<p>Article</p>\n<p>More speed, more models, more fun.</p>\n<p>October 22, 2026 inference / serving / NVIDIA / AMD / TPU / Metal / Intel</p>\n<p>We are pleased to announce the release of <code>zml/llmd 20261023.0</code>.</p>\n<p>While last week’s release (<code>zml/llmd 20261015.0</code>) was a minor release with a focus on performance, this week’s release builds on the gains we made and brings more speed, more models, and more fun.</p>\n<p>And so this week, we bring new model series to zml/llmd.</p>\n<h2>Qwen/Qwen3.8-Flash-Next</h2>\n<p>This release brings the <a target=\"_blank\" rel=\"noopener noreferrer\" href=\"https://huggingface.co/collections/Qwen/qwen38-flash-next\">Qwen/Qwen3.8-Flash-Next series</a> model series to zml/llmd.</p>\n<p>We have tested it on NVIDIA and AMD GPU, the only ones we got access to that could host the official release. Google is still telling us there are no TPUs available.</p>\n<p>Decode throughput, Qwen/Qwen3.8-Flash-Next, bs=16, zml/llmd 20261023.0 (tok/s)</p>\n<h2>XiaomiMiMo/MiMo-V2.6</h2>\n<p>This release brings the <a target=\"_blank\" rel=\"noopener noreferrer\" href=\"https://huggingface.co/collections/XiaomiMiMo/mimo-v26\">XiaomiMiMo/MiMo-V2.6 series</a> in both RL and MOPD flavour.</p>\n<p>Decode throughput, XiaomiMiMo/MiMo-V2.6-Flash-MOPD, bs=16, zml/llmd 20261023.0 (tok/s)</p>\n<h2>Faster DeepSeek-V4.1-Flash</h2>\n<p>The hill climb continues. We improved a number of fundamental building blocks and thus DeepSeek-V4.1-Flash got faster.</p>\n<p>Token Generation, deepseek-ai/DeepSeek-V4.1-Flash, bs=1, 4xGB300 (tok/s)</p>\n<h2>Try it today</h2>\n<p><strong>NVIDIA</strong></p>\n<pre><code class=\"language-bash\"><span class=\"code-snippet\" data-language=\"Shell\"><span class=\"syntax-source syntax-shell syntax-bash\"><span class=\"syntax-meta syntax-function-call syntax-shell\"><span class=\"syntax-variable syntax-function syntax-shell\">docker</span></span><span class=\"syntax-meta syntax-function-call syntax-arguments syntax-shell\"> run<span class=\"syntax-variable syntax-parameter syntax-option syntax-shell\"><span class=\"syntax-punctuation syntax-definition syntax-parameter syntax-shell\"> -</span>p</span> 8000:8000<span class=\"syntax-variable syntax-parameter syntax-option syntax-shell\"><span class=\"syntax-punctuation syntax-definition syntax-parameter syntax-shell\"> --</span>shm-size</span><span class=\"syntax-keyword syntax-operator syntax-assignment syntax-option syntax-shell\">=</span>256GB<span class=\"syntax-variable syntax-parameter syntax-option syntax-shell\"><span class=\"syntax-punctuation syntax-definition syntax-parameter syntax-shell\"> --</span>gpus</span><span class=\"syntax-keyword syntax-operator syntax-assignment syntax-option syntax-shell\">=</span>all<span class=\"syntax-variable syntax-parameter syntax-option syntax-shell\"><span class=\"syntax-punctuation syntax-definition syntax-parameter syntax-shell\"> -</span>e</span> HF_TOKEN<span class=\"syntax-variable syntax-parameter syntax-option syntax-shell\"><span class=\"syntax-punctuation syntax-definition syntax-parameter syntax-shell\"> -</span>it</span> zmlai/llmd:cuda <span class=\"syntax-punctuation syntax-separator syntax-continuation syntax-line syntax-shell\">\\\n</span><span class=\"syntax-variable syntax-parameter syntax-option syntax-shell\"><span class=\"syntax-punctuation syntax-definition syntax-parameter syntax-shell\">    --</span>model</span><span class=\"syntax-keyword syntax-operator syntax-assignment syntax-option syntax-shell\">=</span>hf://Qwen/Qwen3.8-Flash-Next</span>\n</span></span></code></pre>\n<p><strong>AMD</strong></p>\n<pre><code class=\"language-bash\"><span class=\"code-snippet\" data-language=\"Shell\"><span class=\"syntax-source syntax-shell syntax-bash\"><span class=\"syntax-meta syntax-function-call syntax-shell\"><span class=\"syntax-variable syntax-function syntax-shell\">docker</span></span><span class=\"syntax-meta syntax-function-call syntax-arguments syntax-shell\"> run<span class=\"syntax-variable syntax-parameter syntax-option syntax-shell\"><span class=\"syntax-punctuation syntax-definition syntax-parameter syntax-shell\"> -</span>p</span> 8000:8000<span class=\"syntax-variable syntax-parameter syntax-option syntax-shell\"><span class=\"syntax-punctuation syntax-definition syntax-parameter syntax-shell\"> --</span>shm-size</span><span class=\"syntax-keyword syntax-operator syntax-assignment syntax-option syntax-shell\">=</span>256GB<span class=\"syntax-variable syntax-parameter syntax-option syntax-shell\"><span class=\"syntax-punctuation syntax-definition syntax-parameter syntax-shell\"> --</span>device</span><span class=\"syntax-keyword syntax-operator syntax-assignment syntax-option syntax-shell\">=</span>/dev/kfd<span class=\"syntax-variable syntax-parameter syntax-option syntax-shell\"><span class=\"syntax-punctuation syntax-definition syntax-parameter syntax-shell\"> --</span>device</span><span class=\"syntax-keyword syntax-operator syntax-assignment syntax-option syntax-shell\">=</span>/dev/dri<span class=\"syntax-variable syntax-parameter syntax-option syntax-shell\"><span class=\"syntax-punctuation syntax-definition syntax-parameter syntax-shell\"> -</span>e</span> HF_TOKEN<span class=\"syntax-variable syntax-parameter syntax-option syntax-shell\"><span class=\"syntax-punctuation syntax-definition syntax-parameter syntax-shell\"> -</span>it</span> zmlai/llmd:rocm <span class=\"syntax-punctuation syntax-separator syntax-continuation syntax-line syntax-shell\">\\\n</span><span class=\"syntax-variable syntax-parameter syntax-option syntax-shell\"><span class=\"syntax-punctuation syntax-definition syntax-parameter syntax-shell\">    --</span>model</span><span class=\"syntax-keyword syntax-operator syntax-assignment syntax-option syntax-shell\">=</span>hf://Qwen/Qwen3.8-Flash-Next</span>\n</span></span></code></pre>\n<p><strong>Intel</strong></p>\n<pre><code class=\"language-bash\"><span class=\"code-snippet\" data-language=\"Shell\"><span class=\"syntax-source syntax-shell syntax-bash\"><span class=\"syntax-meta syntax-function-call syntax-shell\"><span class=\"syntax-variable syntax-function syntax-shell\">docker</span></span><span class=\"syntax-meta syntax-function-call syntax-arguments syntax-shell\"> run<span class=\"syntax-variable syntax-parameter syntax-option syntax-shell\"><span class=\"syntax-punctuation syntax-definition syntax-parameter syntax-shell\"> -</span>p</span> 8000:8000<span class=\"syntax-variable syntax-parameter syntax-option syntax-shell\"><span class=\"syntax-punctuation syntax-definition syntax-parameter syntax-shell\"> --</span>device</span><span class=\"syntax-keyword syntax-operator syntax-assignment syntax-option syntax-shell\">=</span>/dev/dri<span class=\"syntax-variable syntax-parameter syntax-option syntax-shell\"><span class=\"syntax-punctuation syntax-definition syntax-parameter syntax-shell\"> -</span>e</span> HF_TOKEN<span class=\"syntax-variable syntax-parameter syntax-option syntax-shell\"><span class=\"syntax-punctuation syntax-definition syntax-parameter syntax-shell\"> -</span>it</span> zmlai/llmd:oneapi <span class=\"syntax-punctuation syntax-separator syntax-continuation syntax-line syntax-shell\">\\\n</span><span class=\"syntax-variable syntax-parameter syntax-option syntax-shell\"><span class=\"syntax-punctuation syntax-definition syntax-parameter syntax-shell\">    --</span>model</span><span class=\"syntax-keyword syntax-operator syntax-assignment syntax-option syntax-shell\">=</span>hf://Qwen/Qwen3.8-27B</span>\n</span></span></code></pre>\n<p><strong>Google TPU</strong></p>\n<pre><code class=\"language-bash\"><span class=\"code-snippet\" data-language=\"Shell\"><span class=\"syntax-source syntax-shell syntax-bash\"><span class=\"syntax-meta syntax-function-call syntax-shell\"><span class=\"syntax-variable syntax-function syntax-shell\">docker</span></span><span class=\"syntax-meta syntax-function-call syntax-arguments syntax-shell\"> run<span class=\"syntax-variable syntax-parameter syntax-option syntax-shell\"><span class=\"syntax-punctuation syntax-definition syntax-parameter syntax-shell\"> --</span>net</span><span class=\"syntax-keyword syntax-operator syntax-assignment syntax-option syntax-shell\">=</span>host<span class=\"syntax-variable syntax-parameter syntax-option syntax-shell\"><span class=\"syntax-punctuation syntax-definition syntax-parameter syntax-shell\"> --</span>privileged</span><span class=\"syntax-variable syntax-parameter syntax-option syntax-shell\"><span class=\"syntax-punctuation syntax-definition syntax-parameter syntax-shell\"> -</span>e</span> HF_TOKEN<span class=\"syntax-variable syntax-parameter syntax-option syntax-shell\"><span class=\"syntax-punctuation syntax-definition syntax-parameter syntax-shell\"> -</span>it</span> zmlai/llmd:tpu <span class=\"syntax-punctuation syntax-separator syntax-continuation syntax-line syntax-shell\">\\\n</span><span class=\"syntax-variable syntax-parameter syntax-option syntax-shell\"><span class=\"syntax-punctuation syntax-definition syntax-parameter syntax-shell\">    --</span>model</span><span class=\"syntax-keyword syntax-operator syntax-assignment syntax-option syntax-shell\">=</span>hf://Qwen/Qwen3.8-27B</span>\n</span></span></code></pre>\n<p><strong>Apple Metal</strong></p>\n<pre><code class=\"language-bash\"><span class=\"code-snippet\" data-language=\"Shell\"><span class=\"syntax-source syntax-shell syntax-bash\"><span class=\"syntax-meta syntax-function-call syntax-shell\"><span class=\"syntax-variable syntax-function syntax-shell\">brew</span></span><span class=\"syntax-meta syntax-function-call syntax-arguments syntax-shell\"> install zml/zml/llmd</span>\n<span class=\"syntax-meta syntax-function-call syntax-shell\"><span class=\"syntax-variable syntax-function syntax-shell\">llmd</span></span><span class=\"syntax-meta syntax-function-call syntax-arguments syntax-shell\"><span class=\"syntax-variable syntax-parameter syntax-option syntax-shell\"><span class=\"syntax-punctuation syntax-definition syntax-parameter syntax-shell\"> --</span>model</span><span class=\"syntax-keyword syntax-operator syntax-assignment syntax-option syntax-shell\">=</span>hf://nvidia/Qwen3.8-27B-NVFP4<span class=\"syntax-variable syntax-parameter syntax-option syntax-shell\"><span class=\"syntax-punctuation syntax-definition syntax-parameter syntax-shell\"> --</span>batch-size</span><span class=\"syntax-keyword syntax-operator syntax-assignment syntax-option syntax-shell\">=</span>1</span>\n</span></span></code></pre>\n",
  "content_markdown": "[Back to posts](https://zml.ai/posts/)\n\nArticle\n\nMore speed, more models, more fun.\n\nOctober 22, 2026 inference / serving / NVIDIA / AMD / TPU / Metal / Intel\n\nWe are pleased to announce the release of `zml/llmd 20261023.0`.\n\nWhile last week’s release (`zml/llmd 20261015.0`) was a minor release with a focus on performance, this week’s release builds on the gains we made and brings more speed, more models, and more fun.\n\nAnd so this week, we bring new model series to zml/llmd.\n\n## Qwen/Qwen3.8-Flash-Next\n\nThis release brings the [Qwen/Qwen3.8-Flash-Next series](https://huggingface.co/collections/Qwen/qwen38-flash-next) model series to zml/llmd.\n\nWe have tested it on NVIDIA and AMD GPU, the only ones we got access to that could host the official release. Google is still telling us there are no TPUs available.\n\nDecode throughput, Qwen/Qwen3.8-Flash-Next, bs=16, zml/llmd 20261023.0 (tok/s)\n\n## XiaomiMiMo/MiMo-V2.6\n\nThis release brings the [XiaomiMiMo/MiMo-V2.6 series](https://huggingface.co/collections/XiaomiMiMo/mimo-v26) in both RL and MOPD flavour.\n\nDecode throughput, XiaomiMiMo/MiMo-V2.6-Flash-MOPD, bs=16, zml/llmd 20261023.0 (tok/s)\n\n## Faster DeepSeek-V4.1-Flash\n\nThe hill climb continues. We improved a number of fundamental building blocks and thus DeepSeek-V4.1-Flash got faster.\n\nToken Generation, deepseek-ai/DeepSeek-V4.1-Flash, bs=1, 4xGB300 (tok/s)\n\n## Try it today\n\n**NVIDIA**\n\n```bash\ndocker run -p 8000:8000 --shm-size=256GB --gpus=all -e HF_TOKEN -it zmlai/llmd:cuda \\\n    --model=hf://Qwen/Qwen3.8-Flash-Next\n```\n\n**AMD**\n\n```bash\ndocker run -p 8000:8000 --shm-size=256GB --device=/dev/kfd --device=/dev/dri -e HF_TOKEN -it zmlai/llmd:rocm \\\n    --model=hf://Qwen/Qwen3.8-Flash-Next\n```\n\n**Intel**\n\n```bash\ndocker run -p 8000:8000 --device=/dev/dri -e HF_TOKEN -it zmlai/llmd:oneapi \\\n    --model=hf://Qwen/Qwen3.8-27B\n```\n\n**Google TPU**\n\n```bash\ndocker run --net=host --privileged -e HF_TOKEN -it zmlai/llmd:tpu \\\n    --model=hf://Qwen/Qwen3.8-27B\n```\n\n**Apple Metal**\n\n```bash\nbrew install zml/zml/llmd\nllmd --model=hf://nvidia/Qwen3.8-27B-NVFP4 --batch-size=1\n```\n",
  "dateCreated": "2026-10-23T18:40:08.262074691+00:00",
  "date_created": "2026-10-22T07:00:00+00:00",
  "date_modified": "2026-10-22T07:00:00+00:00",
  "date_published": "2026-10-22T07:00:00+00:00",
  "description": "We are pleased to announce the release of zml/llmd 20261023.0. While last week’s release (zml/llmd 20261015.0) was a minor release with a focus on performance, this week’s release builds on the gains we made and brings more speed, more…",
  "external_url": "https://zml.ai/posts/llmd-20261023.0/",
  "git": {
    "commit": "20068632b65d2830a2d54f25ca1d90d697bea2c0",
    "path": "items/zml-ai/2026/10/2026-10-22-zml-llmd-20261023-0.md",
    "permalink": "https://github.com/aymericbeaumet/aggr-instance/blob/20068632b65d2830a2d54f25ca1d90d697bea2c0/items/zml-ai/2026/10/2026-10-22-zml-llmd-20261023-0.md"
  },
  "id": "urn:aggr:item:e62736fde04b68419e2ff805ad5c596c830004f2",
  "isBasedOn": {
    "@id": "https://zml.ai/posts/llmd-20261023.0/"
  },
  "mainEntity": {
    "@id": "https://zml.ai/posts/llmd-20261023.0/",
    "@type": "CreativeWork",
    "archivedAt": {
      "@id": "https://aggr.aymericbeaumet.com/items/zml-ai/2026-10-22-zml-llmd-20261023-0/#webpage"
    },
    "dateModified": "2026-10-22T07:00:00+00:00",
    "datePublished": "2026-10-22T07:00:00+00:00",
    "headline": "zml/llmd 20261023.0",
    "image": {
      "@type": "ImageObject",
      "caption": "zml/llmd 20261023.0 — More speed, more models, more fun.",
      "height": 90,
      "url": "https://zml.ai/og/background_hu_32943cbdd2e1726c.png",
      "width": 160
    },
    "timeRequired": "PT2M",
    "url": "https://zml.ai/posts/llmd-20261023.0/",
    "wordCount": 310
  },
  "name": "Archived snapshot: zml/llmd 20261023.0",
  "preview": {
    "alt": "zml/llmd 20261023.0 — More speed, more models, more fun.",
    "color": null,
    "height": 90,
    "placeholder": {
      "data_url": "data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAQAAAC1HAwCAAAAC0lEQVR42mNk+A8AAQUBAScY42YAAAAASUVORK5CYII=",
      "hash": ""
    },
    "url": "https://zml.ai/og/background_hu_32943cbdd2e1726c.png",
    "width": 160
  },
  "reading_minutes": 2,
  "replicated_at": null,
  "source": "zml-ai",
  "source_name": "ZML - Model to Metal",
  "summary": "We are pleased to announce the release of zml/llmd 20261023.0. While last week’s release (zml/llmd 20261015.0) was a minor release with a focus on performance, this week’s release builds on the gains we made and brings more speed, more models, and more fun. And so this week, we bring new model series to zml/llmd. Qwen/Qwen3.8-Flash-Next This release brings the Qwen/Qwen3.8-Flash-Next series model series to zml/llmd. We have tested it on NVIDIA and AMD GPU, the only ones we got access to that could host the official release. Google is still telling us there are no TPUs available.",
  "tags": [],
  "temporalCoverage": "2026-10-23T18:40:08.262074691+00:00",
  "title": "zml/llmd 20261023.0",
  "url": "https://aggr.aymericbeaumet.com/items/zml-ai/2026-10-22-zml-llmd-20261023-0/",
  "word_count": 310
}