{"id":53143,"date":"2026-10-04T09:00:00","date_gmt":"2026-10-04T07:00:00","guid":{"rendered":"https:\/\/www.cloudmagazin.com\/?p=53143"},"modified":"2026-10-06T23:02:48","modified_gmt":"2026-10-06T21:02:48","slug":"kolibri-a-german-ai-model-for-your-own-data-center","status":"publish","type":"post","link":"https:\/\/www.cloudmagazin.com\/en\/2026\/10\/04\/kolibri-a-german-ai-model-for-your-own-data-center\/","title":{"rendered":"Kolibri: A German AI Model for Your Own Data Center"},"content":{"rendered":"<p><strong>On 3 October, Aleph Alpha released Kolibri, a language model built in Germany with 78 billion parameters. The weights are published under Apache 2.0 and run on your own hardware. For public administration and industry, sovereignty thus becomes a question of your own data center. The license, the origin of the data and the planned merger with Cohere limit that sovereignty.<\/strong><\/p>\n<h2>Key takeaways<\/h2>\n<div style=\"background:#f8fbfd;border:1px solid rgba(11,183,253,0.28);border-radius:8px;padding:22px 26px;margin:16px 0 34px;box-shadow:0 1px 0 rgba(11,183,253,0.08),0 2px 8px rgba(0,0,0,0.03);\">\n<ul style=\"margin:0;padding-left:20px;line-height:1.75;color:#1a2733;\">\n<li style=\"margin-bottom:10px;\"><strong>Kolibri runs under Apache 2.0 on your own hardware.<\/strong> Of the 78 billion parameters, 3.46 billion are active per token, and in FP8 the model takes up around 78 GB.<\/li>\n<li style=\"margin-bottom:10px;\"><strong>The license covers only weights and configuration.<\/strong> Code, architecture and training procedures are excluded, and operation requires the aleph-alpha-inference package.<\/li>\n<li><strong>The synthetic training data also comes from GLM and Qwen.<\/strong> The vendor itself faces the merger with Cohere, which is still waiting for regulatory approvals.<\/li>\n<\/ul>\n<\/div>\n<p style=\"font-size:0.88em;color:#666;margin:20px 0 32px 0;border-top:1px solid #e5e5e5;border-bottom:1px solid #e5e5e5;padding:10px 0;\"><span style=\"font-family:'SF Mono','Monaco','Consolas',monospace;color:#0879aa;font-weight:700;text-transform:uppercase;font-size:0.72em;letter-spacing:0.14em;margin-right:14px;\">Related:<\/span><a href=\"https:\/\/www.cloudmagazin.com\/en\/2026\/07\/13\/soofi-s-sovereign-doesnt-mean-superior\/\" style=\"color:#333;text-decoration:underline;\">Soofi S: Sovereign Doesn\u2019t Mean Superior<\/a>&nbsp;&nbsp;<span style=\"color:#ccc;\">\/<\/span>&nbsp;&nbsp;<a href=\"https:\/\/www.cloudmagazin.com\/en\/2026\/07\/15\/aws-sovereign-cloud-what-is-truly-separate\/\" style=\"color:#333;text-decoration:underline;\">AWS Sovereign Cloud: What is truly separate<\/a><\/p>\n<h2>A German model for self-hosting<\/h2>\n<p>On 3 October, Aleph Alpha published Kolibri 1 on <a href=\"https:\/\/huggingface.co\/Aleph-Alpha\/Kolibri-1\">Hugging Face<\/a>. The model has 78 billion parameters and works with 3.46 billion of them per token. The Heidelberg-based vendor summed up the claim in its <a href=\"https:\/\/x.com\/Aleph__Alpha\/status\/2106306840657297814\">announcement on X<\/a>: the weights now belong to the users and can be run under Apache 2.0 on their own hardware.<\/p>\n<p>Kolibri was built in Germany and <a href=\"https:\/\/aleph-alpha.com\/en\/blog\/kolibri-has-landed-a-sovereign-open-weight-model\/\">trained<\/a> on infrastructure in Germany and Finland with 768 B200 GPUs. According to the announcement, 21.3 percent of the pre-training tokens are German, while the model card gives a slightly higher share for the pre-training corpus.<\/p>\n<div style=\"background:#004a59;border:1px solid rgba(11,183,253,0.4);border-radius:10px;padding:34px 26px;margin:36px 0;text-align:center;box-shadow:0 0 24px rgba(11,183,253,0.08);\">\n<div style=\"font-family:'SF Mono','Monaco','Consolas',monospace;font-size:11px;color:#4ec9fe;letter-spacing:0.14em;text-transform:uppercase;margin-bottom:10px;\">\/\/ Metric<\/div>\n<div style=\"font-size:48px;font-weight:700;color:#fff;letter-spacing:-0.03em;line-height:1;\">78 GB<\/div>\n<div style=\"font-size:14px;color:rgba(255,255,255,0.85);margin:12px auto 0;max-width:460px;line-height:1.5;\">Kolibri&#8217;s memory footprint in FP8, minimum a single H200, B200 or B300<\/div>\n<div style=\"font-size:12px;color:#4ec9fe;margin-top:12px;\">\/\/ Source: Aleph Alpha, model card<\/div>\n<\/div>\n<p>Pre-training covered 20,000 billion tokens, and across all training phases the total came to almost 24,000 billion.<\/p>\n<p>Aleph Alpha names sovereign, mission-critical work in regulated sectors as its field of use, including public administration, industry, and aerospace. In self-hosted operation, requests and data can stay in your own data center.<\/p>\n<p><strong>What is an open-weight model?<\/strong> An open-weight model is a trained AI model whose weights the vendor releases for download. Operators can therefore run it on their own servers, and requests do not pass through a cloud interface run by the vendor. Whether training code, data and procedures are also open depends on the individual release.<\/p>\n<h2>What self-hosting requires<\/h2>\n<p>In FP8, Kolibri takes up around 78 GB of memory. As a minimum, Aleph Alpha names two A100 cards in the 80 GB variant, two H100 SXM5, or a single H200, B200 or B300. That means the model fits on a single current GPU server.<\/p>\n<p>Kolibri is trained for a context of 262,144 tokens. The vendor has validated the model up to 1,048,576 tokens, but for efficiency and complex tasks it recommends at most the trained length.<\/p>\n<p>First measurements from the <a href=\"https:\/\/forums.developer.nvidia.com\/t\/kolibri-1-aleph-alphas-78b-fp8-moe-runs-on-the-spark-setup-numbers-and-a-few-things-id-love-to-share\/385048\">Nvidia developer forum<\/a> show that Kolibri also runs on a workstation system. On a single DGX Spark, one user reached around 20 tokens per second when decoding with standard vLLM, a second user around 48 with an optimized build. Both are individual measurements with no connection to the vendor.<\/p>\n<h2>License and origin of the data<\/h2>\n<p>For Kolibri, Apache 2.0 applies to the weights and the configuration files. Explicitly excluded are the underlying code, the model architecture, parameter settings and every training procedure. For operation, Aleph Alpha requires its own aleph-alpha-inference package.<\/p>\n<p>Operators therefore hold the weights under a free license regardless of who owns the vendor in the future. For ongoing operation, they remain dependent on its inference package.<\/p>\n<p>For synthetic training data, Aleph Alpha used models from other vendors. According to the <a href=\"https:\/\/aleph-alpha.com\/downloads\/tech-report.pdf\">tech report<\/a>, GLM-5.2 and GLM-5.3 from Z.ai as well as Qwen3.8-27B from Alibaba&#8217;s Qwen team are the main models with which the company generated synthetic data and regenerated parts of open datasets.<\/p>\n<p>Wharton professor Ethan Mollick picked this up on the day of release. Europe&#8217;s new sovereign model had been fine-tuned with data that GLM and Qwen had generated, <a href=\"https:\/\/x.com\/emollick\/status\/2106487816460910912\">he wrote on X<\/a>. Anyone who also measures sovereignty by the origin of the training data will find the limit of that promise here.<\/p>\n<div style=\"display:grid;grid-template-columns:repeat(auto-fit,minmax(280px,1fr));gap:16px;margin:28px 0;\">\n<div style=\"background:#f7fbf9;border:1px solid rgba(45,122,62,0.22);border-radius:8px;padding:20px 22px;box-shadow:0 2px 8px rgba(0,0,0,0.03);\">\n<div style=\"font-family:'SF Mono','Monaco','Consolas',monospace;font-size:11px;color:#2d7a3e;letter-spacing:0.14em;text-transform:uppercase;margin-bottom:12px;\">\/\/ supports<\/div>\n<ul style=\"margin:0;padding-left:18px;color:#1a2733;line-height:1.6;font-size:0.95em;\">\n<li style=\"margin-bottom:7px;\">Weights and configuration under Apache 2.0<\/li>\n<li style=\"margin-bottom:7px;\">Training in Germany and Finland<\/li>\n<li style=\"margin-bottom:7px;\">Around one fifth of pre-training tokens in German<\/li>\n<li>Operation from a single H200, B200 or B300<\/li>\n<\/ul>\n<\/div>\n<div style=\"background:#fbf8f6;border:1px solid rgba(181,83,42,0.22);border-radius:8px;padding:20px 22px;box-shadow:0 2px 8px rgba(0,0,0,0.03);\">\n<div style=\"font-family:'SF Mono','Monaco','Consolas',monospace;font-size:11px;color:#b5532a;letter-spacing:0.14em;text-transform:uppercase;margin-bottom:12px;\">\/\/ open<\/div>\n<ul style=\"margin:0;padding-left:18px;color:#1a2733;line-height:1.6;font-size:0.95em;\">\n<li style=\"margin-bottom:7px;\">Code, architecture and training procedures not licensed<\/li>\n<li style=\"margin-bottom:7px;\">Operation only with aleph-alpha-inference<\/li>\n<li style=\"margin-bottom:7px;\">Synthetic data from GLM and Qwen<\/li>\n<li>Merger with Cohere waiting for approvals<\/li>\n<\/ul>\n<\/div>\n<\/div>\n<h2>How Kolibri performs in German<\/h2>\n<p>In the vendor&#8217;s table, Kolibri scores 70.8 points in the German overall value, placing it just ahead of GPT-OSS 120B and Qwen3.5 35B-A3B.<\/p>\n<div style=\"overflow-x:auto;margin:24px 0 6px;-webkit-overflow-scrolling:touch;border:1px solid rgba(11,183,253,0.25);border-radius:10px;\">\n<table style=\"border-collapse:collapse;width:100%;font-size:0.88em;\">\n<thead>\n<tr style=\"background:#f8fbfd;\">\n<th style=\"text-align:left;padding:13px 14px;border-bottom:2px solid #0bb7fd;color:#1a2733;position:sticky;left:0;background:#f8fbfd;z-index:2;\">Model<\/th>\n<th style=\"text-align:center;padding:13px 10px;border-bottom:2px solid #0bb7fd;background:rgba(11,183,253,0.14);color:#1a2733;font-weight:700;\">German overall score (Overall DE)<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td style=\"text-align:left;padding:12px 14px;border-bottom:1px solid #eef2f5;position:sticky;left:0;background:#fff;font-weight:600;color:#1a2733;\">Qwen3.8 27B (dense model)<\/td>\n<td style=\"text-align:center;padding:12px 10px;border-bottom:1px solid #eef2f5;color:#333;\">79.9<\/td>\n<\/tr>\n<tr>\n<td style=\"text-align:left;padding:12px 14px;border-bottom:1px solid #eef2f5;position:sticky;left:0;background:#fff;font-weight:600;color:#1a2733;\">Kolibri 1<\/td>\n<td style=\"text-align:center;padding:12px 10px;border-bottom:1px solid #eef2f5;background:rgba(11,183,253,0.08);font-weight:700;color:#0879aa;\">70.8<\/td>\n<\/tr>\n<tr>\n<td style=\"text-align:left;padding:12px 14px;border-bottom:1px solid #eef2f5;position:sticky;left:0;background:#fff;font-weight:600;color:#1a2733;\">GPT-OSS 120B<\/td>\n<td style=\"text-align:center;padding:12px 10px;border-bottom:1px solid #eef2f5;color:#333;\">70.2<\/td>\n<\/tr>\n<tr>\n<td style=\"text-align:left;padding:12px 14px;position:sticky;left:0;background:#fff;font-weight:600;color:#1a2733;\">Qwen3.5 35B-A3B<\/td>\n<td style=\"text-align:center;padding:12px 10px;color:#333;\">69.8<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p style=\"font-size:0.8em;color:#888;margin:6px 2px 30px;\">Source: Aleph Alpha vendor table, model card on Hugging Face<\/p>\n<\/div>\n<p>The dense model Qwen3.8 27B is clearly ahead with 79.9 points. Per token, however, it activates almost eight times as many parameters as Kolibri.<\/p>\n<p>Aleph Alpha chief Ilhan Scheer announced Kolibri 1.1 on 4 October, <a href=\"https:\/\/x.com\/IlhanScheer\/status\/2106787681875415231\">faster than expected<\/a>. For the internal predecessor model, he cited a gain in German from 46 to 71 points within three months.<\/p>\n<h2>The vendor before the merger<\/h2>\n<p>Kolibri arrives just over two weeks after Aleph Alpha and Cohere signed a <a href=\"https:\/\/www.prnewswire.com\/news-releases\/cohere-and-aleph-alpha-sign-agreement-to-become-the-first-transatlantic-sovereign-ai-solution-302880732.html\">binding merger agreement<\/a> on 16 September. The joint company is to operate as Cohere, based in Berlin and Toronto. The Heidelberg office is to remain as a research center.<\/p>\n<p>The Schwarz Group companies are providing <a href=\"https:\/\/aleph-alpha.com\/news\/cohere-aleph-alpha-zusammenschluss-geplant\/\">500 million euros<\/a> in structured financing for the German-Canadian venture. The transaction only closes once the final regulatory approvals have been granted.<\/p>\n<p>For operators, this shifts the question of who stands behind Kolibri. The Kolibri 1 weights remain usable under Apache 2.0, and operating them requires the vendor&#8217;s inference package. The announced Kolibri 1.1 also comes from a vendor that is to operate under the name Cohere in the future.<\/p>\n<h2 style=\"padding-top:64px;margin-bottom:20px;\">Frequently Asked Questions<\/h2>\n<details>\n<summary><strong>Can Kolibri be used commercially?<\/strong><\/summary>\n<p style=\"margin:8px 0 4px 24px;color:#555;line-height:1.6;\">The weights are published under Apache 2.0, a license that permits commercial use. It covers the weights and the configuration files, while code, architecture and training procedures are excluded. For operation, Aleph Alpha requires the aleph-alpha-inference package.<\/p>\n<\/details>\n<details>\n<summary><strong>What hardware does Kolibri need?<\/strong><\/summary>\n<p style=\"margin:8px 0 4px 24px;color:#555;line-height:1.6;\">In FP8, the model takes up around 78 GB of memory. As a minimum, Aleph Alpha names two A100 cards with 80 GB, two H100 SXM5, or a single H200, B200 or B300. Community measurements show that Kolibri also runs on a DGX Spark, at around 20 to 48 tokens per second.<\/p>\n<\/details>\n<details>\n<summary><strong>What does the merger with Cohere mean for Kolibri users?<\/strong><\/summary>\n<p style=\"margin:8px 0 4px 24px;color:#555;line-height:1.6;\">The agreement of 16 September is signed, but it is only completed after the final regulatory approvals. The published weights remain usable under Apache 2.0. Running Kolibri 1 requires the aleph-alpha-inference package, which the vendor provides.<\/p>\n<\/details>\n<h2>Editor&#8217;s Picks<\/h2>\n<ul style=\"list-style:none;margin:0 0 8px;padding:0;\">\n<li style=\"padding:10px 0;border-bottom:1px solid #eee;\"><a href=\"https:\/\/www.cloudmagazin.com\/en\/2026\/08\/04\/downloadable-doesnt-mean-deployable\/\" style=\"color:#1a1a1a;text-decoration:none;\">Downloadable Doesn\u2019t Mean Deployable<\/a><\/li>\n<li style=\"padding:10px 0;border-bottom:1px solid #eee;\"><a href=\"https:\/\/www.cloudmagazin.com\/en\/2026\/09\/06\/nvidia-acquires-hugging-face-open-models-remain-an-option\/\" style=\"color:#1a1a1a;text-decoration:none;\">Nvidia Acquires Hugging Face: Open Models Remain an Option<\/a><\/li>\n<li style=\"padding:10px 0;\"><a href=\"https:\/\/www.cloudmagazin.com\/en\/2026\/05\/20\/finops-for-ai-inference-getting-a-grip-on-gpu-costs-in-multi-cloud\/\" style=\"color:#1a1a1a;text-decoration:none;\">FinOps for AI Inference: Managing GPU Costs in Multi-Cloud<\/a><\/li>\n<\/ul>\n<p style=\"margin:44px 0 12px;font-size:0.72em;font-weight:700;text-transform:uppercase;letter-spacing:0.18em;color:#666;\">More from the MBF Media Network<\/p>\n<div style=\"padding:14px 18px;border-left:3px solid #202528;background:#fafafa;\">\n<div style=\"font-size:0.7em;font-weight:700;color:#202528;text-transform:uppercase;letter-spacing:0.12em;margin-bottom:4px;\">mybusinessfuture<\/div>\n<p><a href=\"https:\/\/mybusinessfuture.com\/en\/when-a-german-ai-model-truly-pays-off\/\" style=\"font-weight:600;line-height:1.4;color:#1a1a1a;text-decoration:none;\">When a German AI Model Truly Pays Off<\/a><\/p>\n<\/div>\n<p style=\"text-align:right;color:#868e96;font-size:0.75em;margin-top:40px;\"><em>Image source: AI-generated (October 2026)<\/em><\/p>\n","protected":false},"excerpt":{"rendered":"Kolibri runs on your own hardware. Its license, training data and pending merger with Cohere set limits on sovereignty.","protected":false},"author":87,"featured_media":53138,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"_yoast_wpseo_meta-robots-noindex":"","_yoast_wpseo_meta-robots-nofollow":"","_yoast_wpseo_meta-robots-adv":"","_yoast_wpseo_canonical":"","_yoast_wpseo_opengraph-title":"","_yoast_wpseo_opengraph-description":"","_yoast_wpseo_opengraph-image":"https:\/\/www.cloudmagazin.com\/wp-content\/uploads\/2026\/10\/cover-hero-3.jpg","_yoast_wpseo_opengraph-image-id":53138,"_yoast_wpseo_twitter-title":"","_yoast_wpseo_twitter-description":"","_yoast_wpseo_twitter-image":"https:\/\/www.cloudmagazin.com\/wp-content\/uploads\/2026\/10\/cover-hero-3.jpg","_yoast_wpseo_twitter-image-id":53138,"pre_headline":"","bildquelle":"","teasertext":"Kolibri runs on your own hardware. Its license, training data and pending merger with Cohere set limits on sovereignty.","language":"en","_evm_slot_owner":"","_evm_translation_lang":"en","featured_post":0,"featured_post_sortierung":0,"_wp_old_slug":[],"footnotes":""},"categories":[930],"tags":[],"industry":[],"class_list":["post-53143","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-reboot-germany"],"evm_reading_time_minutes":7,"wpml_language":"en","wpml_translation_of":53127,"acf":[],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v27.9 - https:\/\/yoast.com\/product\/yoast-seo-wordpress\/ -->\n<title>Kolibri: A German AI Model for Your Own Data Center<\/title>\n<meta name=\"description\" content=\"Kolibri\u2019s open weights run on your own hardware. Apache 2.0 covers weights and configuration; data origins and the merger limit sovereignty.\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/www.cloudmagazin.com\/en\/2026\/10\/04\/kolibri-a-german-ai-model-for-your-own-data-center\/\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Kolibri: A German AI Model for Your Own Data Center\" \/>\n<meta property=\"og:description\" content=\"Kolibri\u2019s open weights run on your own hardware. Apache 2.0 covers weights and configuration; data origins and the merger limit sovereignty.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/www.cloudmagazin.com\/en\/2026\/10\/04\/kolibri-a-german-ai-model-for-your-own-data-center\/\" \/>\n<meta property=\"og:site_name\" content=\"cloudmagazin\" \/>\n<meta property=\"article:publisher\" content=\"https:\/\/www.facebook.com\/cloudmagazincom\/\" \/>\n<meta property=\"article:published_time\" content=\"2026-10-04T07:00:00+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2026-10-06T21:02:48+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/www.cloudmagazin.com\/wp-content\/uploads\/2026\/10\/cover-hero-3.jpg\" \/>\n\t<meta property=\"og:image:width\" content=\"1792\" \/>\n\t<meta property=\"og:image:height\" content=\"1024\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/jpeg\" \/>\n<meta name=\"author\" content=\"Benedikt Langer\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:image\" content=\"https:\/\/www.cloudmagazin.com\/wp-content\/uploads\/2026\/10\/cover-hero-3.jpg\" \/>\n<meta name=\"twitter:creator\" content=\"@cloudmagazin\" \/>\n<meta name=\"twitter:site\" content=\"@cloudmagazin\" \/>\n<meta name=\"twitter:label1\" content=\"Written by\" \/>\n\t<meta name=\"twitter:data1\" content=\"Benedikt Langer\" \/>\n\t<meta name=\"twitter:label2\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data2\" content=\"6 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"NewsArticle\",\"@id\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/2026\\\/10\\\/04\\\/kolibri-a-german-ai-model-for-your-own-data-center\\\/#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/2026\\\/10\\\/04\\\/kolibri-a-german-ai-model-for-your-own-data-center\\\/\"},\"author\":{\"name\":\"Benedikt Langer\",\"@id\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/#\\\/schema\\\/person\\\/9275a9f6961b8f365c1548e316b21d19\"},\"headline\":\"Kolibri: A German AI Model for Your Own Data Center\",\"datePublished\":\"2026-10-04T07:00:00+00:00\",\"dateModified\":\"2026-10-06T21:02:48+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/2026\\\/10\\\/04\\\/kolibri-a-german-ai-model-for-your-own-data-center\\\/\"},\"wordCount\":1140,\"publisher\":{\"@id\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/#organization\"},\"image\":{\"@id\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/2026\\\/10\\\/04\\\/kolibri-a-german-ai-model-for-your-own-data-center\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/www.cloudmagazin.com\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/cover-hero-3.jpg\",\"articleSection\":[\"Reboot Germany\"],\"inLanguage\":\"en-US\"},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/2026\\\/10\\\/04\\\/kolibri-a-german-ai-model-for-your-own-data-center\\\/\",\"url\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/2026\\\/10\\\/04\\\/kolibri-a-german-ai-model-for-your-own-data-center\\\/\",\"name\":\"Kolibri: A German AI Model for Your Own Data Center\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/2026\\\/10\\\/04\\\/kolibri-a-german-ai-model-for-your-own-data-center\\\/#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/2026\\\/10\\\/04\\\/kolibri-a-german-ai-model-for-your-own-data-center\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/www.cloudmagazin.com\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/cover-hero-3.jpg\",\"datePublished\":\"2026-10-04T07:00:00+00:00\",\"dateModified\":\"2026-10-06T21:02:48+00:00\",\"description\":\"Kolibri\u2019s open weights run on your own hardware. Apache 2.0 covers weights and configuration; data origins and the merger limit sovereignty.\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/2026\\\/10\\\/04\\\/kolibri-a-german-ai-model-for-your-own-data-center\\\/#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/2026\\\/10\\\/04\\\/kolibri-a-german-ai-model-for-your-own-data-center\\\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/2026\\\/10\\\/04\\\/kolibri-a-german-ai-model-for-your-own-data-center\\\/#primaryimage\",\"url\":\"https:\\\/\\\/www.cloudmagazin.com\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/cover-hero-3.jpg\",\"contentUrl\":\"https:\\\/\\\/www.cloudmagazin.com\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/cover-hero-3.jpg\",\"width\":1792,\"height\":1024,\"caption\":\"Editorial-Collage zu Aleph Alphas offenem Sprachmodell Kolibri: ein Kolibri, dessen K\u00f6rper und Fl\u00fcgel aus Leiterbahnen und Chip-Strukturen b\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/2026\\\/10\\\/04\\\/kolibri-a-german-ai-model-for-your-own-data-center\\\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/home\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Kolibri: A German AI Model for Your Own Data Center\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/#website\",\"url\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/\",\"name\":\"cloudmagazin\",\"description\":\"Inspiration f\u00fcr Businessentscheider\",\"publisher\":{\"@id\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/#organization\"},\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en-US\"},{\"@type\":\"Organization\",\"@id\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/#organization\",\"name\":\"cloudmagazin\",\"url\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/\",\"logo\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/#\\\/schema\\\/logo\\\/image\\\/\",\"url\":\"https:\\\/\\\/www.cloudmagazin.com\\\/wp-content\\\/uploads\\\/2020\\\/04\\\/cloudmagazin-logo-klein_menu.jpg\",\"contentUrl\":\"https:\\\/\\\/www.cloudmagazin.com\\\/wp-content\\\/uploads\\\/2020\\\/04\\\/cloudmagazin-logo-klein_menu.jpg\",\"width\":150,\"height\":150,\"caption\":\"cloudmagazin\"},\"image\":{\"@id\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/#\\\/schema\\\/logo\\\/image\\\/\"},\"sameAs\":[\"https:\\\/\\\/www.facebook.com\\\/cloudmagazincom\\\/\",\"https:\\\/\\\/x.com\\\/cloudmagazin\",\"https:\\\/\\\/www.linkedin.com\\\/showcase\\\/cloudmagazin\\\/\"]},{\"@type\":\"Person\",\"@id\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/#\\\/schema\\\/person\\\/9275a9f6961b8f365c1548e316b21d19\",\"name\":\"Benedikt Langer\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/www.cloudmagazin.com\\\/wp-content\\\/uploads\\\/2026\\\/01\\\/evernine-bilder-benedikt_1.jpg.jpg\",\"url\":\"https:\\\/\\\/www.cloudmagazin.com\\\/wp-content\\\/uploads\\\/2026\\\/01\\\/evernine-bilder-benedikt_1.jpg.jpg\",\"contentUrl\":\"https:\\\/\\\/www.cloudmagazin.com\\\/wp-content\\\/uploads\\\/2026\\\/01\\\/evernine-bilder-benedikt_1.jpg.jpg\",\"caption\":\"Benedikt Langer\"},\"description\":\"Benedikt Langer focuses on IT and cloud topics as an editor, with a particular emphasis on artificial intelligence, digital infrastructure, and strategic cloud architectures. In his articles, he examines technological developments from the perspective of decision-makers, integrating them into economic, regulatory, and organizational contexts. In addition to Cloudmagazin, he regularly contributes to other specialized magazines within Evernine Media.\",\"sameAs\":[\"https:\\\/\\\/www.linkedin.com\\\/in\\\/benedikt-langer\\\/\"],\"url\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/author\\\/benedikt\\\/\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"Kolibri: A German AI Model for Your Own Data Center","description":"Kolibri\u2019s open weights run on your own hardware. Apache 2.0 covers weights and configuration; data origins and the merger limit sovereignty.","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/www.cloudmagazin.com\/en\/2026\/10\/04\/kolibri-a-german-ai-model-for-your-own-data-center\/","og_locale":"en_US","og_type":"article","og_title":"Kolibri: A German AI Model for Your Own Data Center","og_description":"Kolibri\u2019s open weights run on your own hardware. Apache 2.0 covers weights and configuration; data origins and the merger limit sovereignty.","og_url":"https:\/\/www.cloudmagazin.com\/en\/2026\/10\/04\/kolibri-a-german-ai-model-for-your-own-data-center\/","og_site_name":"cloudmagazin","article_publisher":"https:\/\/www.facebook.com\/cloudmagazincom\/","article_published_time":"2026-10-04T07:00:00+00:00","article_modified_time":"2026-10-06T21:02:48+00:00","og_image":[{"width":1792,"height":1024,"url":"https:\/\/www.cloudmagazin.com\/wp-content\/uploads\/2026\/10\/cover-hero-3.jpg","type":"image\/jpeg"}],"author":"Benedikt Langer","twitter_card":"summary_large_image","twitter_image":"https:\/\/www.cloudmagazin.com\/wp-content\/uploads\/2026\/10\/cover-hero-3.jpg","twitter_creator":"@cloudmagazin","twitter_site":"@cloudmagazin","twitter_misc":{"Written by":"Benedikt Langer","Est. reading time":"6 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"NewsArticle","@id":"https:\/\/www.cloudmagazin.com\/en\/2026\/10\/04\/kolibri-a-german-ai-model-for-your-own-data-center\/#article","isPartOf":{"@id":"https:\/\/www.cloudmagazin.com\/en\/2026\/10\/04\/kolibri-a-german-ai-model-for-your-own-data-center\/"},"author":{"name":"Benedikt Langer","@id":"https:\/\/www.cloudmagazin.com\/en\/#\/schema\/person\/9275a9f6961b8f365c1548e316b21d19"},"headline":"Kolibri: A German AI Model for Your Own Data Center","datePublished":"2026-10-04T07:00:00+00:00","dateModified":"2026-10-06T21:02:48+00:00","mainEntityOfPage":{"@id":"https:\/\/www.cloudmagazin.com\/en\/2026\/10\/04\/kolibri-a-german-ai-model-for-your-own-data-center\/"},"wordCount":1140,"publisher":{"@id":"https:\/\/www.cloudmagazin.com\/en\/#organization"},"image":{"@id":"https:\/\/www.cloudmagazin.com\/en\/2026\/10\/04\/kolibri-a-german-ai-model-for-your-own-data-center\/#primaryimage"},"thumbnailUrl":"https:\/\/www.cloudmagazin.com\/wp-content\/uploads\/2026\/10\/cover-hero-3.jpg","articleSection":["Reboot Germany"],"inLanguage":"en-US"},{"@type":"WebPage","@id":"https:\/\/www.cloudmagazin.com\/en\/2026\/10\/04\/kolibri-a-german-ai-model-for-your-own-data-center\/","url":"https:\/\/www.cloudmagazin.com\/en\/2026\/10\/04\/kolibri-a-german-ai-model-for-your-own-data-center\/","name":"Kolibri: A German AI Model for Your Own Data Center","isPartOf":{"@id":"https:\/\/www.cloudmagazin.com\/en\/#website"},"primaryImageOfPage":{"@id":"https:\/\/www.cloudmagazin.com\/en\/2026\/10\/04\/kolibri-a-german-ai-model-for-your-own-data-center\/#primaryimage"},"image":{"@id":"https:\/\/www.cloudmagazin.com\/en\/2026\/10\/04\/kolibri-a-german-ai-model-for-your-own-data-center\/#primaryimage"},"thumbnailUrl":"https:\/\/www.cloudmagazin.com\/wp-content\/uploads\/2026\/10\/cover-hero-3.jpg","datePublished":"2026-10-04T07:00:00+00:00","dateModified":"2026-10-06T21:02:48+00:00","description":"Kolibri\u2019s open weights run on your own hardware. Apache 2.0 covers weights and configuration; data origins and the merger limit sovereignty.","breadcrumb":{"@id":"https:\/\/www.cloudmagazin.com\/en\/2026\/10\/04\/kolibri-a-german-ai-model-for-your-own-data-center\/#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/www.cloudmagazin.com\/en\/2026\/10\/04\/kolibri-a-german-ai-model-for-your-own-data-center\/"]}]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/www.cloudmagazin.com\/en\/2026\/10\/04\/kolibri-a-german-ai-model-for-your-own-data-center\/#primaryimage","url":"https:\/\/www.cloudmagazin.com\/wp-content\/uploads\/2026\/10\/cover-hero-3.jpg","contentUrl":"https:\/\/www.cloudmagazin.com\/wp-content\/uploads\/2026\/10\/cover-hero-3.jpg","width":1792,"height":1024,"caption":"Editorial-Collage zu Aleph Alphas offenem Sprachmodell Kolibri: ein Kolibri, dessen K\u00f6rper und Fl\u00fcgel aus Leiterbahnen und Chip-Strukturen b"},{"@type":"BreadcrumbList","@id":"https:\/\/www.cloudmagazin.com\/en\/2026\/10\/04\/kolibri-a-german-ai-model-for-your-own-data-center\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/www.cloudmagazin.com\/en\/home\/"},{"@type":"ListItem","position":2,"name":"Kolibri: A German AI Model for Your Own Data Center"}]},{"@type":"WebSite","@id":"https:\/\/www.cloudmagazin.com\/en\/#website","url":"https:\/\/www.cloudmagazin.com\/en\/","name":"cloudmagazin","description":"Inspiration f\u00fcr Businessentscheider","publisher":{"@id":"https:\/\/www.cloudmagazin.com\/en\/#organization"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/www.cloudmagazin.com\/en\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-US"},{"@type":"Organization","@id":"https:\/\/www.cloudmagazin.com\/en\/#organization","name":"cloudmagazin","url":"https:\/\/www.cloudmagazin.com\/en\/","logo":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/www.cloudmagazin.com\/en\/#\/schema\/logo\/image\/","url":"https:\/\/www.cloudmagazin.com\/wp-content\/uploads\/2020\/04\/cloudmagazin-logo-klein_menu.jpg","contentUrl":"https:\/\/www.cloudmagazin.com\/wp-content\/uploads\/2020\/04\/cloudmagazin-logo-klein_menu.jpg","width":150,"height":150,"caption":"cloudmagazin"},"image":{"@id":"https:\/\/www.cloudmagazin.com\/en\/#\/schema\/logo\/image\/"},"sameAs":["https:\/\/www.facebook.com\/cloudmagazincom\/","https:\/\/x.com\/cloudmagazin","https:\/\/www.linkedin.com\/showcase\/cloudmagazin\/"]},{"@type":"Person","@id":"https:\/\/www.cloudmagazin.com\/en\/#\/schema\/person\/9275a9f6961b8f365c1548e316b21d19","name":"Benedikt Langer","image":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/www.cloudmagazin.com\/wp-content\/uploads\/2026\/01\/evernine-bilder-benedikt_1.jpg.jpg","url":"https:\/\/www.cloudmagazin.com\/wp-content\/uploads\/2026\/01\/evernine-bilder-benedikt_1.jpg.jpg","contentUrl":"https:\/\/www.cloudmagazin.com\/wp-content\/uploads\/2026\/01\/evernine-bilder-benedikt_1.jpg.jpg","caption":"Benedikt Langer"},"description":"Benedikt Langer focuses on IT and cloud topics as an editor, with a particular emphasis on artificial intelligence, digital infrastructure, and strategic cloud architectures. In his articles, he examines technological developments from the perspective of decision-makers, integrating them into economic, regulatory, and organizational contexts. In addition to Cloudmagazin, he regularly contributes to other specialized magazines within Evernine Media.","sameAs":["https:\/\/www.linkedin.com\/in\/benedikt-langer\/"],"url":"https:\/\/www.cloudmagazin.com\/en\/author\/benedikt\/"}]}},"_links":{"self":[{"href":"https:\/\/www.cloudmagazin.com\/en\/wp-json\/wp\/v2\/posts\/53143","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.cloudmagazin.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.cloudmagazin.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.cloudmagazin.com\/en\/wp-json\/wp\/v2\/users\/87"}],"replies":[{"embeddable":true,"href":"https:\/\/www.cloudmagazin.com\/en\/wp-json\/wp\/v2\/comments?post=53143"}],"version-history":[{"count":5,"href":"https:\/\/www.cloudmagazin.com\/en\/wp-json\/wp\/v2\/posts\/53143\/revisions"}],"predecessor-version":[{"id":53244,"href":"https:\/\/www.cloudmagazin.com\/en\/wp-json\/wp\/v2\/posts\/53143\/revisions\/53244"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.cloudmagazin.com\/en\/wp-json\/wp\/v2\/media\/53138"}],"wp:attachment":[{"href":"https:\/\/www.cloudmagazin.com\/en\/wp-json\/wp\/v2\/media?parent=53143"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.cloudmagazin.com\/en\/wp-json\/wp\/v2\/categories?post=53143"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.cloudmagazin.com\/en\/wp-json\/wp\/v2\/tags?post=53143"},{"taxonomy":"industry","embeddable":true,"href":"https:\/\/www.cloudmagazin.com\/en\/wp-json\/wp\/v2\/industry?post=53143"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}