{"id":49065,"date":"2026-07-13T10:59:31","date_gmt":"2026-07-13T08:59:31","guid":{"rendered":"https:\/\/www.cloudmagazin.com\/2026\/07\/13\/soofi-s-sovereign-doesnt-mean-superior\/"},"modified":"2026-07-13T18:30:48","modified_gmt":"2026-07-13T16:30:48","slug":"soofi-s-sovereign-doesnt-mean-superior","status":"publish","type":"post","link":"https:\/\/www.cloudmagazin.com\/en\/2026\/07\/13\/soofi-s-sovereign-doesnt-mean-superior\/","title":{"rendered":"Soofi S: Sovereign Doesn\u2019t Mean Superior"},"content":{"rendered":"<p style=\"line-height:1.8;margin-bottom:20px;\"><strong>Soofi S 30B-A3B achieves the highest scores on German benchmarks among current European models. Its hybrid MoE-Mamba architecture delivers higher decode throughput at 40K context than pure attention models of the same size class. For an infrastructure team, the decision to deploy ultimately comes down to one pragmatic question: Do these strengths outweigh the beta status and the lack of operational processes?<\/strong><\/p>\n<h2>Key Points in Brief<\/h2>\n<div style=\"background:#eaf6fc;border-left:4px solid #0bb7fd;border-radius:0 8px 8px 0;padding:22px 26px;margin:28px 0 34px;\">\n<ul style=\"margin:0;padding-left:22px;line-height:1.7;color:#12333f;\">\n<li style=\"margin-bottom:10px;\"><strong>Strong in German, weak internationally.<\/strong> According to project data, Soofi leads in the German aggregate (79.1) and German code, but clearly loses to Qwen3.5 in English and reasoning.<\/li>\n<li style=\"margin-bottom:10px;\"><strong>Architecture is the lever.<\/strong> The Mamba component keeps decode throughput high with long context, where full attention models break down due to the KV cache.<\/li>\n<li style=\"margin-bottom:0;\"><strong>Beta without SLA.<\/strong> There is no commercial support. Patches, monitoring, and outages are entirely the operator\u2019s responsibility.<\/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=\"color:#00004c;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\/09\/fast-at-top-level-cheaper-and-trained-across-three-continents\/\" style=\"color:#333;text-decoration:underline;\">Almost at top level, trained more cheaply<\/a>&nbsp;&nbsp;<span style=\"color:#ccc;\">\/<\/span>&nbsp;&nbsp;<a href=\"https:\/\/www.cloudmagazin.com\/en\/2026\/07\/12\/sovereign-cloud-does-not-end-at-the-server-location\/\" style=\"color:#333;text-decoration:underline;\">Sovereign Cloud doesn\u2019t end at the server location<\/a><\/p>\n<h2 style=\"margin-top:64px;margin-bottom:20px;padding-top:16px;\">Performance Comparison with Qwen and European Models<\/h2>\n<p style=\"line-height:1.8;margin-bottom:20px;\">Soofi scores 79.1 points on the German aggregates, ahead of Nemotron at 74.9. The gap widens on German code. MBPP-DE reaches 84.2, marking the best value among the compared models there. These figures are manufacturer-provided and based on training data with strong German weighting.<\/p>\n<p style=\"line-height:1.8;margin-bottom:20px;\">Qwen3.5 35B outperforms Soofi on the English aggregate, 74.6 to 70.1. The deficit is equally clear on reasoning benchmarks such as GPQA and BBH. Workloads with a high share of English or abstract reasoning tasks therefore show a noticeable disadvantage versus the Chinese model. The table lists the direct comparison values across all test fields.<\/p>\n<style>\n  .soofi-mod{--cm:#0bb7fd;--soofi:#7b3fe4;--soofi-bg:#efe7fb;--nemo-bg:#e3f3e6;\n    --ink:#1a1f26;--mut:#5c6672;--line:#e5e8ec;--bg:#ffffff;\n    font-family:-apple-system,BlinkMacSystemFont,\"Segoe UI\",Roboto,Helvetica,Arial,sans-serif;\n    color:var(--ink);background:var(--bg);max-width:900px;margin:2rem auto;padding:0 4px;-webkit-font-smoothing:antialiased}\n  .soofi-mod *{box-sizing:border-box}\n  .soofi-mod .kicker{font-size:.72rem;letter-spacing:.14em;text-transform:uppercase;color:var(--cm);font-weight:700;margin:0 0 .35rem}\n  .soofi-mod h3{font-size:1.28rem;line-height:1.25;margin:0 0 .3rem;font-weight:750}\n  .soofi-mod .sub{font-size:.9rem;color:var(--mut);margin:0 0 1.1rem;line-height:1.45}\n  .soofi-mod .tbl-wrap{overflow-x:auto;border:1px solid var(--line);border-radius:12px}\n  .soofi-mod table{border-collapse:collapse;width:100%;min-width:640px;font-size:.83rem}\n  .soofi-mod thead th{position:sticky;top:0;background:#f7f8fa;text-align:right;padding:.7rem .6rem;\n    font-weight:700;color:var(--ink);border-bottom:2px solid var(--line);vertical-align:bottom;white-space:nowrap}\n  .soofi-mod thead th:first-child{text-align:left}\n  .soofi-mod thead th.c-soofi{background:var(--soofi-bg);color:var(--soofi)}\n  .soofi-mod thead th.c-nemo{background:var(--nemo-bg)}\n  .soofi-mod th .mdl{display:block;font-size:.82em;font-weight:600;color:var(--mut)}\n  .soofi-mod thead th.c-soofi .mdl{color:var(--soofi);opacity:.85}\n  .soofi-mod tbody th{text-align:left;font-weight:500;padding:.42rem .6rem;color:var(--ink);white-space:nowrap}\n  .soofi-mod tbody td{text-align:right;padding:.42rem .6rem;font-variant-numeric:tabular-nums}\n  .soofi-mod td.c-soofi{background:var(--soofi-bg)}\n  .soofi-mod td.c-nemo{background:var(--nemo-bg)}\n  .soofi-mod tr.grp th{background:#fbfcfd;font-weight:750;font-size:.8rem;letter-spacing:.02em;\n    text-transform:uppercase;color:var(--mut);padding-top:.7rem;padding-bottom:.35rem;border-top:1px solid var(--line)}\n  .soofi-mod b.win{font-weight:800;color:var(--ink)}\n  .soofi-mod u.sec{text-decoration:underline;text-decoration-color:#b9c0c9;text-underline-offset:2px}\n  .soofi-mod .legend{font-size:.76rem;color:var(--mut);margin:.7rem 2px 0;line-height:1.5}\n  .soofi-mod .legend b{color:var(--ink)}\n  .soofi-mod .chip{display:inline-block;width:.7em;height:.7em;border-radius:3px;vertical-align:-1px;margin-right:.25em}\n  .soofi-mod .src{font-size:.74rem;color:var(--mut);font-style:italic;text-align:right;margin:.55rem 2px 0}\n  \/* Chart *\/\n  .soofi-mod .chart-card{border:1px solid var(--line);border-radius:12px;padding:1.1rem 1rem .8rem;margin-top:2.4rem;background:var(--bg)}\n  .soofi-mod svg{width:100%;height:auto;display:block}\n  .soofi-mod .c-legend{display:flex;flex-wrap:wrap;gap:.4rem 1.1rem;font-size:.78rem;color:var(--mut);margin:.7rem 4px 0}\n  .soofi-mod .c-legend span{display:inline-flex;align-items:center;gap:.35em}\n  .soofi-mod details.soofi-more{margin-top:12px}\n  .soofi-mod details.soofi-more>summary{cursor:pointer;list-style:none;font-size:.85rem;font-weight:700;color:var(--cm);padding:11px 15px;border:1px solid var(--line);border-radius:8px;display:flex;align-items:center;gap:.5em;-webkit-user-select:none;user-select:none;background:var(--bg)}\n  .soofi-mod details.soofi-more>summary::-webkit-details-marker{display:none}\n  .soofi-mod details.soofi-more>summary::before{content:\"\\25B8\";display:inline-block;transition:transform .2s}\n  .soofi-mod details.soofi-more[open]>summary::before{transform:rotate(90deg)}\n  .soofi-mod details.soofi-more>summary::after{content:none!important;display:none!important;background:none!important}\n  @media (max-width:600px){\n    .soofi-mod table{font-size:.74rem;min-width:520px}\n    .soofi-mod thead th{padding:.5rem .38rem}\n    .soofi-mod tbody th,.soofi-mod tbody td{padding:.34rem .38rem}\n    .soofi-mod h3{font-size:1.12rem}\n    .soofi-mod .chart-card{padding:.8rem .5rem .7rem}\n  }\n<\/style>\n<div class=\"soofi-mod\" style=\"margin:2rem auto\">\n<!-- ========== BLOCK 1: BENCHMARK-TABELLE ========== --><\/p>\n<p class=\"kicker\">Benchmark in Peer Comparison<\/p>\n<h3>Soofi S 30B-A3B versus Open Competition<\/h3>\n<p class=\"sub\">Five model families, aggregated test fields plus German specialist benchmarks. Bold = best value in the row, underlined = second-best. Soofi leads primarily where German matters.<\/p>\n<div class=\"tbl-wrap\" style=\"overflow-x:auto;-webkit-overflow-scrolling:touch;\">\n<table style=\"min-width:560px;\">\n<thead>\n<tr>\n<th>Benchmark<\/th>\n<th class=\"c-soofi\">Soofi S<span class=\"mdl\">30B-A3B<\/span><\/th>\n<th class=\"c-nemo\">Nemotron 3<span class=\"mdl\">Nano 30B-A3B<\/span><\/th>\n<th>Qwen3.5<span class=\"mdl\">35B-A3B<\/span><\/th>\n<th>Ministral 3<span class=\"mdl\">14B<\/span><\/th>\n<th>Gemma 3<span class=\"mdl\">27B<\/span><\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr class=\"grp\">\n<th colspan=\"6\">Aggregates<\/th>\n<\/tr>\n<tr>\n<th>English aggregate<\/th>\n<td class=\"c-soofi\">70.1<\/td>\n<td class=\"c-nemo\">68.3<\/td>\n<td><b class=\"win\">74.6<\/b><\/td>\n<td><u class=\"sec\">70.3<\/u><\/td>\n<td><u class=\"sec\">70.3<\/u><\/td>\n<\/tr>\n<tr>\n<th>German aggregate<\/th>\n<td class=\"c-soofi\"><u class=\"sec\">79.1<\/u><\/td>\n<td class=\"c-nemo\">74.9<\/td>\n<td><b class=\"win\">81.6<\/b><\/td>\n<td>78.3<\/td>\n<td>78.4<\/td>\n<\/tr>\n<tr>\n<th>Held-out (EN)<\/th>\n<td class=\"c-soofi\"><u class=\"sec\">41.4<\/u><\/td>\n<td class=\"c-nemo\">34.7<\/td>\n<td><b class=\"win\">47.8<\/b><\/td>\n<td>40.0<\/td>\n<td>35.5<\/td>\n<\/tr>\n<tr>\n<th>Held-out (DE)<\/th>\n<td class=\"c-soofi\">41.8<\/td>\n<td class=\"c-nemo\">39.8<\/td>\n<td><b class=\"win\">46.0<\/b><\/td>\n<td><u class=\"sec\">42.2<\/u><\/td>\n<td>38.3<\/td>\n<\/tr>\n<\/tbody>\n<\/table><\/div>\n<details class=\"soofi-more\">\n<summary>Show all individual benchmarks (code, mathematics, knowledge, reasoning and more)<\/summary>\n<div class=\"tbl-wrap\" style=\"overflow-x:auto;-webkit-overflow-scrolling:touch;margin-top:12px;\">\n<table style=\"min-width:560px;\">\n<thead>\n<tr>\n<th>Benchmark<\/th>\n<th class=\"c-soofi\">Soofi S<span class=\"mdl\">30B-A3B<\/span><\/th>\n<th class=\"c-nemo\">Nemotron 3<span class=\"mdl\">Nano 30B-A3B<\/span><\/th>\n<th>Qwen3.5<span class=\"mdl\">35B-A3B<\/span><\/th>\n<th>Ministral 3<span class=\"mdl\">14B<\/span><\/th>\n<th>Gemma 3<span class=\"mdl\">27B<\/span><\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr class=\"grp\">\n<th colspan=\"6\">Code (pass@1)<\/th>\n<\/tr>\n<tr>\n<th>HumanEval<\/th>\n<td class=\"c-soofi\"><b class=\"win\">73.8<\/b><\/td>\n<td class=\"c-nemo\"><u class=\"sec\">72.2<\/u><\/td>\n<td>67.1<\/td>\n<td>60.9<\/td>\n<td>60.2<\/td>\n<\/tr>\n<tr>\n<th>MBPP<\/th>\n<td class=\"c-soofi\"><b class=\"win\">70.2<\/b><\/td>\n<td class=\"c-nemo\"><u class=\"sec\">67.5<\/u><\/td>\n<td>65.8<\/td>\n<td>58.8<\/td>\n<td>66.7<\/td>\n<\/tr>\n<tr>\n<th>LBPP<\/th>\n<td class=\"c-soofi\">31.0<\/td>\n<td class=\"c-nemo\"><b class=\"win\">38.1<\/b><\/td>\n<td><u class=\"sec\">32.4<\/u><\/td>\n<td>27.0<\/td>\n<td>22.4<\/td>\n<\/tr>\n<tr>\n<th>HumanEval-DE<\/th>\n<td class=\"c-soofi\"><u class=\"sec\">65.5<\/u><\/td>\n<td class=\"c-nemo\"><b class=\"win\">68.8<\/b><\/td>\n<td>59.5<\/td>\n<td>55.5<\/td>\n<td>57.7<\/td>\n<\/tr>\n<tr>\n<th>MBPP-DE<\/th>\n<td class=\"c-soofi\"><b class=\"win\">84.2<\/b><\/td>\n<td class=\"c-nemo\"><u class=\"sec\">79.9<\/u><\/td>\n<td>79.0<\/td>\n<td>72.0<\/td>\n<td>75.6<\/td>\n<\/tr>\n<tr class=\"grp\">\n<th colspan=\"6\">Mathematics<\/th>\n<\/tr>\n<tr>\n<th>GSM8K<\/th>\n<td class=\"c-soofi\"><u class=\"sec\">86.1<\/u><\/td>\n<td class=\"c-nemo\"><b class=\"win\">86.5<\/b><\/td>\n<td>82.6<\/td>\n<td>83.7<\/td>\n<td>79.9<\/td>\n<\/tr>\n<tr>\n<th>GSM8K-Platinum-DE<\/th>\n<td class=\"c-soofi\">87.1<\/td>\n<td class=\"c-nemo\"><u class=\"sec\">87.7<\/u><\/td>\n<td><b class=\"win\">91.2<\/b><\/td>\n<td>81.3<\/td>\n<td>80.7<\/td>\n<\/tr>\n<tr>\n<th>Minerva 500<\/th>\n<td class=\"c-soofi\"><u class=\"sec\">79.4<\/u><\/td>\n<td class=\"c-nemo\">64.0<\/td>\n<td><b class=\"win\">83.0<\/b><\/td>\n<td>61.8<\/td>\n<td>75.7<\/td>\n<\/tr>\n<tr>\n<th>Minerva Math-EN<\/th>\n<td class=\"c-soofi\"><u class=\"sec\">81.0<\/u><\/td>\n<td class=\"c-nemo\">64.2<\/td>\n<td><b class=\"win\">82.4<\/b><\/td>\n<td>61.3<\/td>\n<td>73.7<\/td>\n<\/tr>\n<tr>\n<th>Minerva MATH-DE<\/th>\n<td class=\"c-soofi\">56.0<\/td>\n<td class=\"c-nemo\">58.1<\/td>\n<td><b class=\"win\">76.5<\/b><\/td>\n<td>55.1<\/td>\n<td><u class=\"sec\">65.6<\/u><\/td>\n<\/tr>\n<tr class=\"grp\">\n<th colspan=\"6\">Knowledge<\/th>\n<\/tr>\n<tr>\n<th>MMLU stem<\/th>\n<td class=\"c-soofi\">75.9<\/td>\n<td class=\"c-nemo\">74.8<\/td>\n<td><b class=\"win\">81.9<\/b><\/td>\n<td><u class=\"sec\">76.4<\/u><\/td>\n<td>74.7<\/td>\n<\/tr>\n<tr>\n<th>MMLU-Pro<\/th>\n<td class=\"c-soofi\">51.4<\/td>\n<td class=\"c-nemo\">51.6<\/td>\n<td><b class=\"win\">60.3<\/b><\/td>\n<td><u class=\"sec\">53.3<\/u><\/td>\n<td>50.7<\/td>\n<\/tr>\n<tr>\n<th>MMLU-Pro-DE<\/th>\n<td class=\"c-soofi\">49.4<\/td>\n<td class=\"c-nemo\">47.1<\/td>\n<td><b class=\"win\">56.6<\/b><\/td>\n<td><u class=\"sec\">50.2<\/u><\/td>\n<td>47.2<\/td>\n<\/tr>\n<tr>\n<th>INCLUDE-DE<\/th>\n<td class=\"c-soofi\"><b class=\"win\">61.2<\/b><\/td>\n<td class=\"c-nemo\">59.7<\/td>\n<td>61.2<\/td>\n<td><u class=\"sec\">60.4<\/u><\/td>\n<td>57.6<\/td>\n<\/tr>\n<tr>\n<th>NaturalQuestions (acc)<\/th>\n<td class=\"c-soofi\">79.0<\/td>\n<td class=\"c-nemo\">80.3<\/td>\n<td><u class=\"sec\">81.0<\/u><\/td>\n<td>77.1<\/td>\n<td><b class=\"win\">83.5<\/b><\/td>\n<\/tr>\n<tr class=\"grp\">\n<th colspan=\"6\">Commonsense &amp; Reading Comprehension<\/th>\n<\/tr>\n<tr>\n<th>PIQA<\/th>\n<td class=\"c-soofi\"><b class=\"win\">85.7<\/b><\/td>\n<td class=\"c-nemo\">84.8<\/td>\n<td><u class=\"sec\">84.9<\/u><\/td>\n<td>82.9<\/td>\n<td><u class=\"sec\">84.9<\/u><\/td>\n<\/tr>\n<tr>\n<th>PIQA-DE<\/th>\n<td class=\"c-soofi\"><b class=\"win\">91.5<\/b><\/td>\n<td class=\"c-nemo\">85.7<\/td>\n<td><u class=\"sec\">87.6<\/u><\/td>\n<td>81.3<\/td>\n<td>86.7<\/td>\n<\/tr>\n<tr>\n<th>SocialIQA<\/th>\n<td class=\"c-soofi\"><b class=\"win\">60.5<\/b><\/td>\n<td class=\"c-nemo\">57.6<\/td>\n<td><u class=\"sec\">58.0<\/u><\/td>\n<td>53.9<\/td>\n<td>56.2<\/td>\n<\/tr>\n<tr>\n<th>SocialIQA-DE<\/th>\n<td class=\"c-soofi\"><b class=\"win\">87.6<\/b><\/td>\n<td class=\"c-nemo\">82.7<\/td>\n<td><u class=\"sec\">87.2<\/u><\/td>\n<td>83.0<\/td>\n<td>86.7<\/td>\n<\/tr>\n<tr>\n<th>SQuAD (EM)<\/th>\n<td class=\"c-soofi\"><b class=\"win\">87.5<\/b><\/td>\n<td class=\"c-nemo\"><u class=\"sec\">87.3<\/u><\/td>\n<td>80.4<\/td>\n<td>82.7<\/td>\n<td>85.1<\/td>\n<\/tr>\n<tr>\n<th>DROP (EM)<\/th>\n<td class=\"c-soofi\"><b class=\"win\">66.5<\/b><\/td>\n<td class=\"c-nemo\">64.2<\/td>\n<td>62.6<\/td>\n<td>62.9<\/td>\n<td><u class=\"sec\">65.0<\/u><\/td>\n<\/tr>\n<tr class=\"grp\">\n<th colspan=\"6\">Reasoning &amp; Natural Sciences<\/th>\n<\/tr>\n<tr>\n<th>BBH (CoT)<\/th>\n<td class=\"c-soofi\">78.8<\/td>\n<td class=\"c-nemo\">77.8<\/td>\n<td><b class=\"win\">84.2<\/b><\/td>\n<td><u class=\"sec\">79.6<\/u><\/td>\n<td>77.8<\/td>\n<\/tr>\n<tr>\n<th>AGIEval<\/th>\n<td class=\"c-soofi\">66.9<\/td>\n<td class=\"c-nemo\">65.1<\/td>\n<td><b class=\"win\">71.7<\/b><\/td>\n<td>67.0<\/td>\n<td><u class=\"sec\">68.3<\/u><\/td>\n<\/tr>\n<tr>\n<th>GPQA-Diamond<\/th>\n<td class=\"c-soofi\"><u class=\"sec\">43.4<\/u><\/td>\n<td class=\"c-nemo\">33.8<\/td>\n<td><b class=\"win\">50.0<\/b><\/td>\n<td>40.9<\/td>\n<td>35.9<\/td>\n<\/tr>\n<tr>\n<th>GPQA-Diamond-DE<\/th>\n<td class=\"c-soofi\">41.9<\/td>\n<td class=\"c-nemo\">37.4<\/td>\n<td><b class=\"win\">46.4<\/b><\/td>\n<td><u class=\"sec\">42.5<\/u><\/td>\n<td>31.8<\/td>\n<\/tr>\n<tr>\n<th>ARC-Challenge<\/th>\n<td class=\"c-soofi\">90.6<\/td>\n<td class=\"c-nemo\">89.4<\/td>\n<td><b class=\"win\">93.1<\/b><\/td>\n<td><u class=\"sec\">91.6<\/u><\/td>\n<td><u class=\"sec\">91.6<\/u><\/td>\n<\/tr>\n<tr class=\"grp\">\n<th colspan=\"6\">German Language Proficiency<\/th>\n<\/tr>\n<tr>\n<th>GLP-DE<\/th>\n<td class=\"c-soofi\">88.8<\/td>\n<td class=\"c-nemo\">73.7<\/td>\n<td><b class=\"win\">94.0<\/b><\/td>\n<td><u class=\"sec\">89.5<\/u><\/td>\n<td>88.3<\/td>\n<\/tr>\n<tr>\n<th>ARC-Challenge-DE<\/th>\n<td class=\"c-soofi\">92.3<\/td>\n<td class=\"c-nemo\">91.4<\/td>\n<td><b class=\"win\">96.6<\/b><\/td>\n<td><u class=\"sec\">93.5<\/u><\/td>\n<td>93.3<\/td>\n<\/tr>\n<\/tbody>\n<\/table><\/div>\n<\/details>\n<p class=\"legend\">\n    <span class=\"chip\" style=\"background:#efe7fb\"><\/span><b>Soofi S<\/b> &#8211; the German consortium model &nbsp;&nbsp;<br \/>\n    <span class=\"chip\" style=\"background:#e3f3e6\"><\/span><b>Nemotron 3 Nano<\/b> &#8211; the closest architectural reference &nbsp;&nbsp;<br \/>\n    <b>Bold<\/b> best, <u class=\"sec\">underlined<\/u> second-best value per row.\n  <\/p>\n<p class=\"src\">Source: Soofi Pretraining Tech Report (Soofi-Project)<\/p>\n<\/div>\n<p style=\"line-height:1.8;margin-bottom:20px;\">The European models Teuken-7B, Apertus, EuroLLM and Salamandra lag further behind in the aggregates. Teuken-7B sits at the bottom left of the chart. Aleph Alpha was acquired by Cohere in April 2026. According to the benchmarks presented, Soofi is currently the strongest open European model of this scale that is not under non-European corporate ownership.<\/p>\n<h2 style=\"margin-top:64px;margin-bottom:20px;padding-top:16px;\">What the Hybrid MoE-Mamba Design Means in Practice<\/h2>\n<p style=\"line-height:1.8;margin-bottom:20px;\">Soofi uses a total of 30 billion parameters but activates only around three billion per token. The Mamba component replaces parts of the classic attention mechanism and scales linearly with context length. This keeps decode speed in tokens per second per GPU high even at 40K context. Full attention models struggle here with the growing KV cache, which consumes memory bandwidth and reduces possible batch sizes. The Mamba component instead maintains a constant state and does not require a KV cache.<\/p>\n<p style=\"line-height:1.8;margin-bottom:20px;\">For an infrastructure team, this translates into lower latency with long documents, extended chat histories, or retrieval contexts. A key distinction matters in practice and can prove costly if overlooked: the active parameter share reduces compute per token, not weight storage. All 30 billion parameters remain in VRAM; savings occur in FLOPs and activation memory. An additional cost arises from the MoE routing logic. It introduces overhead for batching, load balancing, and scheduling expert activations, plus communication overhead in multi-GPU operation. The diagram positions Soofi according to capability and measured throughput.<\/p>\n<div class=\"soofi-mod\" style=\"margin:2rem auto\">\n<!-- ========== BLOCK 2: CAPABILITY-INDEX vs DECODE-SPEED ========== --><\/p>\n<div class=\"chart-card\">\n<p class=\"kicker\" style=\"margin-bottom:.3rem\">Capability vs. Speed<\/p>\n<h3 style=\"margin-bottom:.25rem\">Capability Index vs. Decode Speed at 40K Context<\/h3>\n<p class=\"sub\" style=\"margin-bottom:.9rem\">The further right, the faster per GPU. The further up, the more capable. Soofi S sits in the top right &#8211; a position no European model has occupied until now.<\/p>\n<p>    <img decoding=\"async\" src=\"https:\/\/www.cloudmagazin.com\/wp-content\/uploads\/2026\/07\/soofi-capability-vs-speed-de.png\" alt=\"Scatter plot: Capability index versus aggregated decode speed per GPU at 40K context. Soofi S 30B-A3B sits top right, at highest capability and highest throughput.\" style=\"width:100%;height:auto;display:block;border-radius:6px\" \/><\/p>\n<div class=\"c-legend\">\n      <span><span class=\"chip\" style=\"background:#a020f0\"><\/span>Soofi S<\/span><br \/>\n      <span><span class=\"chip\" style=\"background:#3b6bd6;border-radius:50%\"><\/span>Open European models<\/span><br \/>\n      <span><span class=\"chip\" style=\"background:#9aa2ad;border-radius:50%\"><\/span>International models<\/span><br \/>\n      <span><span class=\"chip\" style=\"background:#e8a33d\"><\/span>Open non-EU model<\/span><br \/>\n      <span><span class=\"chip\" style=\"background:#5c6672\"><\/span>Dense reference<\/span>\n    <\/div>\n<p class=\"src\">Recreation based on Soofi Pretraining Tech Report &#8211; values rounded, logarithmic axis<\/p>\n<\/p><\/div>\n<\/div>\n<p style=\"line-height:1.8;margin-bottom:20px;\">It shows the model in the top right, at high capability and the highest measured throughput. No European model has been in this field until now. The position results from the combination of MoE selectivity and Mamba efficiency with long context.<\/p>\n<h2 style=\"margin-top:64px;margin-bottom:20px;padding-top:16px;\">Resource Requirements and What Is Realistic Today<\/h2>\n<p style=\"line-height:1.8;margin-bottom:20px;\">According to the manufacturer, training consumed approximately 253,000 B200 GPU hours on Deutsche Telekom\u2019s Industrial AI Cloud in Munich. Around 27,000 billion training tokens with a strong German weighting were used. The license is permissive, but access is currently restricted to a closed beta.<\/p>\n<div style=\"background:#004a59;color:#fff;border-radius:8px;padding:26px 30px;margin:30px 0;\">\n<p style=\"margin:0 0 6px;font-family:monospace;font-size:0.78em;letter-spacing:0.1em;color:#7fdcf0;\">\/\/ active per token<\/p>\n<p style=\"margin:0;font-size:2.6em;font-weight:800;line-height:1;letter-spacing:-0.02em;color:#ffffff;\">3 billion<\/p>\n<p style=\"margin:12px 0 0;line-height:1.6;color:rgba(255,255,255,0.85);font-size:0.95em;\">of 30 billion total parameters are actually computed per token. The computational effort per token is therefore closer to that of a 3-billion model. VRAM requirements for the weights, however, remain at the level of the full 30B model.<\/p>\n<\/div>\n<p style=\"line-height:1.8;margin-bottom:20px;\">During inference, the computational effort per token is therefore closer to that of a model with three billion active parameters. A team saves compute time but must still allocate memory for the weights of a 30-billion model. The beta lacks stable versions, documented update paths, and clear information on compatibility with common inference frameworks such as vLLM or TensorRT-LLM.<\/p>\n<p style=\"line-height:1.8;margin-bottom:20px;\">What is realistically possible today are internal feasibility studies, long-context experiments, and initial fine-tuning on proprietary data. For production systems with defined availability and latency SLAs, the current state is usually insufficient. Integration into existing MLOps pipelines requires additional in-house development and validation work.<\/p>\n<h2 style=\"margin-top:64px;margin-bottom:20px;padding-top:16px;\">The Efficiency Debate on Training Effort<\/h2>\n<p style=\"line-height:1.8;margin-bottom:20px;\">Critics in the expert community point to roughly 80 percent architectural and mixture overlap with Nvidia\u2019s Nemotron 3 Nano. In their view, continual pretraining on the existing model would have required far less compute. The 253.000 B200 GPU hours therefore appear excessive from this perspective. Some observers also regard the self-defined capability index metric as overstated.<\/p>\n<p style=\"line-height:1.8;margin-bottom:20px;\">The team led by Michael Fromm pushes back: the objective was never the lowest possible eval score. A model trained from scratch with its own data pipeline offers better technical reproducibility and more targeted further development than simply building on an external checkpoint.<\/p>\n<p style=\"line-height:1.8;margin-bottom:20px;\">The technical trade-off is straightforward. Teams that require full control over training data and weighting accept the higher cost. Those looking for a strong baseline with minimal additional effort tend to choose continual pretraining or an off-the-shelf model.<\/p>\n<h2 style=\"margin-top:64px;margin-bottom:20px;padding-top:16px;\">Who is Responsible for Operations<\/h2>\n<p style=\"line-height:1.8;margin-bottom:20px;\">In production, a team manages patches, security updates, and incident response itself. There is no commercial provider that assumes responsibility for SLAs, monitoring, or fixed response times. Any outage or security vulnerability falls entirely on the operator.<\/p>\n<p style=\"line-height:1.8;margin-bottom:20px;\">The training work originates from a research consortium. Further development depends on the continuation of this project. Anyone deploying Soofi creates a dependency on a development process that is not primarily designed for operational stability. This increases the effort required for in-house observability, rollback strategies, and emergency procedures. This exact point remains the toughest open question when moving from benchmark to real-world operation.<\/p>\n<div style=\"background:#eaf6fc;border-left:4px solid #0bb7fd;border-radius:0 8px 8px 0;padding:20px 24px;margin:30px 0;\">\n<p style=\"margin:0;line-height:1.7;color:#12333f;\"><strong>What is a hybrid MoE-Mamba model?<\/strong> Mixture-of-Experts activates only a portion of the parameters per token &#8211; around three out of thirty billion in Soofi\u2019s case. The Mamba component replaces parts of the classic attention mechanism and scales linearly with sequence length. The hybrid design combines selective activation with efficient long-context processing, achieving high decode throughput without the full attention overhead.<\/p>\n<\/div>\n<h2 style=\"padding-top:64px;margin-bottom:20px;\">Frequently Asked Questions<\/h2>\n<details>\n<summary><strong>Which tasks are currently best suited for Soofi?<\/strong><\/summary>\n<p>German specialist content, German code, and long-context tasks with high throughput requirements are a good fit. On English or heavily reasoning-intensive workloads, Qwen3.5 35B leads. The table and diagram in the article show the profile in detail.<\/p>\n<\/details>\n<details>\n<summary><strong>What does the closed beta mean for deployment in an organization?<\/strong><\/summary>\n<p>Access is restricted and there is no guarantee of stable releases or timely fixes. Teams should first test the model in isolated environments and implement their own validation and fallback mechanisms before it enters critical paths.<\/p>\n<\/details>\n<details>\n<summary><strong>Who handles updates and maintenance after training?<\/strong><\/summary>\n<p>The consortium currently does not provide production-ready operational processes. Operators apply patches themselves, manage versions, and respond independently to new releases. This responsibility remains one of the biggest practical hurdles on the path to live operations.<\/p>\n<\/details>\n<h2>Editor&#8217;s Picks<\/h2>\n<ul>\n<li><a href=\"https:\/\/www.cloudmagazin.com\/en\/2026\/07\/09\/fast-at-top-level-cheaper-and-trained-across-three-continents\/\">Nearly at top level, more affordable, and trained across three continents<\/a><\/li>\n<li><a href=\"https:\/\/www.cloudmagazin.com\/en\/2026\/07\/12\/sovereign-cloud-does-not-end-at-the-server-location\/\">Sovereign Cloud doesn&#8217;t end at the server location<\/a><\/li>\n<li><a href=\"https:\/\/www.cloudmagazin.com\/en\/2026\/07\/11\/bsi-c3a-cloud-sovereignty-becomes-auditable\/\">BSI (Federal Office for Information Security) C3A: Cloud sovereignty becomes verifiable<\/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;margin-bottom:6px;\">\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\/wann-sich-deutsches-ki-modell-rechnet\/\" style=\"font-weight:600;line-height:1.4;color:#1a1a1a;text-decoration:none;\">When a German AI model really pays off<\/a>\n<\/div>\n<div style=\"padding:14px 18px;border-left:3px solid #d65663;background:#fafafa;margin-bottom:6px;\">\n<div style=\"font-size:0.7em;font-weight:700;color:#d65663;text-transform:uppercase;letter-spacing:0.12em;margin-bottom:4px;\">digital-chiefs<\/div>\n<p><a href=\"https:\/\/www.digital-chiefs.de\/souveraene-ki-verantwortung-bleibt-im-haus\/\" style=\"font-weight:600;line-height:1.4;color:#1a1a1a;text-decoration:none;\">Sovereign AI: responsibility stays in-house<\/a>\n<\/div>\n<div style=\"padding:14px 18px;border-left:3px solid #69d8ed;background:#fafafa;margin-bottom:6px;\">\n<div style=\"font-size:0.7em;font-weight:700;color:#69d8ed;text-transform:uppercase;letter-spacing:0.12em;margin-bottom:4px;\">securitytoday<\/div>\n<p><a href=\"https:\/\/www.securitytoday.de\/en\/2026\/07\/07\/der-ai-act-ist-in-wahrheit-ein-security-gesetz\/\" style=\"font-weight:600;line-height:1.4;color:#1a1a1a;text-decoration:none;\">The AI Act is actually a security law<\/a>\n<\/div>\n<p style=\"text-align:right;margin:28px 0 8px;\"><em style=\"font-size:0.82em;color:#888;\">Source cover image: AI-generated (July 2026)<\/em><\/p>\n","protected":false},"excerpt":{"rendered":"Soofi S leads in German benchmarks and maintains speed in long-context tasks, but Qwen3.5 outperforms in reasoning. A technical performance review.","protected":false},"author":31,"featured_media":49030,"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":"","_yoast_wpseo_opengraph-image-id":0,"_yoast_wpseo_twitter-title":"","_yoast_wpseo_twitter-description":"","_yoast_wpseo_twitter-image":"","_yoast_wpseo_twitter-image-id":0,"pre_headline":"","bildquelle":"","teasertext":"","language":"de","_evm_slot_owner":"","_evm_translation_lang":"en","featured_post":0,"featured_post_sortierung":0,"_wp_old_slug":[],"footnotes":""},"categories":[924,930],"tags":[],"industry":[],"class_list":["post-49065","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-artificial-intelligence","category-reboot-germany"],"evm_reading_time_minutes":9,"wpml_language":"en","wpml_translation_of":49028,"acf":[],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v27.9 - https:\/\/yoast.com\/product\/yoast-seo-wordpress\/ -->\n<title>Soofi S: Sovereign Doesn\u2019t Mean Superior - cloudmagazin<\/title>\n<meta name=\"description\" content=\"Soofi S tops German benchmarks and excels in long-context speed, yet Qwen3.5 dominates reasoning. Explore the full technical performance breakdown.\" \/>\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\/07\/13\/soofi-s-sovereign-doesnt-mean-superior\/\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Soofi S: Sovereign Doesn\u2019t Mean Superior - cloudmagazin\" \/>\n<meta property=\"og:description\" content=\"Soofi S tops German benchmarks and excels in long-context speed, yet Qwen3.5 dominates reasoning. Explore the full technical performance breakdown.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/www.cloudmagazin.com\/en\/2026\/07\/13\/soofi-s-sovereign-doesnt-mean-superior\/\" \/>\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=\"2026-07-13T08:59:31+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2026-07-13T16:30:48+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/www.cloudmagazin.com\/wp-content\/uploads\/2026\/07\/soofi-s-souveraen-heisst-nicht-sieger-cover-hero-1.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=\"Alec Chizhik\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\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=\"Alec Chizhik\" \/>\n\t<meta name=\"twitter:label2\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data2\" content=\"7 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\\\/07\\\/13\\\/soofi-s-sovereign-doesnt-mean-superior\\\/#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/2026\\\/07\\\/13\\\/soofi-s-sovereign-doesnt-mean-superior\\\/\"},\"author\":{\"name\":\"Alec Chizhik\",\"@id\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/#\\\/schema\\\/person\\\/ce38baaa19a580268aedce096597eb3c\"},\"headline\":\"Soofi S: Sovereign Doesn\u2019t Mean Superior\",\"datePublished\":\"2026-07-13T08:59:31+00:00\",\"dateModified\":\"2026-07-13T16:30:48+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/2026\\\/07\\\/13\\\/soofi-s-sovereign-doesnt-mean-superior\\\/\"},\"wordCount\":1486,\"publisher\":{\"@id\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/#organization\"},\"image\":{\"@id\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/2026\\\/07\\\/13\\\/soofi-s-sovereign-doesnt-mean-superior\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/www.cloudmagazin.com\\\/wp-content\\\/uploads\\\/2026\\\/07\\\/soofi-s-souveraen-heisst-nicht-sieger-cover-hero-1.jpg\",\"articleSection\":[\"Artificial Intelligence\",\"Reboot Germany\"],\"inLanguage\":\"en-US\"},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/2026\\\/07\\\/13\\\/soofi-s-sovereign-doesnt-mean-superior\\\/\",\"url\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/2026\\\/07\\\/13\\\/soofi-s-sovereign-doesnt-mean-superior\\\/\",\"name\":\"Soofi S: Sovereign Doesn\u2019t Mean Superior - cloudmagazin\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/2026\\\/07\\\/13\\\/soofi-s-sovereign-doesnt-mean-superior\\\/#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/2026\\\/07\\\/13\\\/soofi-s-sovereign-doesnt-mean-superior\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/www.cloudmagazin.com\\\/wp-content\\\/uploads\\\/2026\\\/07\\\/soofi-s-souveraen-heisst-nicht-sieger-cover-hero-1.jpg\",\"datePublished\":\"2026-07-13T08:59:31+00:00\",\"dateModified\":\"2026-07-13T16:30:48+00:00\",\"description\":\"Soofi S tops German benchmarks and excels in long-context speed, yet Qwen3.5 dominates reasoning. Explore the full technical performance breakdown.\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/2026\\\/07\\\/13\\\/soofi-s-sovereign-doesnt-mean-superior\\\/#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/2026\\\/07\\\/13\\\/soofi-s-sovereign-doesnt-mean-superior\\\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/2026\\\/07\\\/13\\\/soofi-s-sovereign-doesnt-mean-superior\\\/#primaryimage\",\"url\":\"https:\\\/\\\/www.cloudmagazin.com\\\/wp-content\\\/uploads\\\/2026\\\/07\\\/soofi-s-souveraen-heisst-nicht-sieger-cover-hero-1.jpg\",\"contentUrl\":\"https:\\\/\\\/www.cloudmagazin.com\\\/wp-content\\\/uploads\\\/2026\\\/07\\\/soofi-s-souveraen-heisst-nicht-sieger-cover-hero-1.jpg\",\"width\":1792,\"height\":1024,\"caption\":\"Parallele Linien und eine t\u00fcrkisfarbene Kante auf hellem Grund.\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/2026\\\/07\\\/13\\\/soofi-s-sovereign-doesnt-mean-superior\\\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/home\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Soofi S: Sovereign Doesn\u2019t Mean Superior\"}]},{\"@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\\\/ce38baaa19a580268aedce096597eb3c\",\"name\":\"Alec Chizhik\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/www.cloudmagazin.com\\\/wp-content\\\/uploads\\\/2026\\\/03\\\/alec-chizhik.jpg\",\"url\":\"https:\\\/\\\/www.cloudmagazin.com\\\/wp-content\\\/uploads\\\/2026\\\/03\\\/alec-chizhik.jpg\",\"contentUrl\":\"https:\\\/\\\/www.cloudmagazin.com\\\/wp-content\\\/uploads\\\/2026\\\/03\\\/alec-chizhik.jpg\",\"caption\":\"Alec Chizhik\"},\"description\":\"Alec is the Chief Digital Officer at Evernine and writes about cloud architectures, IT security, and digital operations practices.\",\"sameAs\":[\"https:\\\/\\\/www.linkedin.com\\\/in\\\/alecchizhik\\\/\"],\"url\":\"https:\\\/\\\/www.cloudmagazin.com\\\/en\\\/author\\\/alec\\\/\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"Soofi S: Sovereign Doesn\u2019t Mean Superior - cloudmagazin","description":"Soofi S tops German benchmarks and excels in long-context speed, yet Qwen3.5 dominates reasoning. Explore the full technical performance breakdown.","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\/07\/13\/soofi-s-sovereign-doesnt-mean-superior\/","og_locale":"en_US","og_type":"article","og_title":"Soofi S: Sovereign Doesn\u2019t Mean Superior - cloudmagazin","og_description":"Soofi S tops German benchmarks and excels in long-context speed, yet Qwen3.5 dominates reasoning. Explore the full technical performance breakdown.","og_url":"https:\/\/www.cloudmagazin.com\/en\/2026\/07\/13\/soofi-s-sovereign-doesnt-mean-superior\/","og_site_name":"cloudmagazin","article_publisher":"https:\/\/www.facebook.com\/cloudmagazincom\/","article_published_time":"2026-07-13T08:59:31+00:00","article_modified_time":"2026-07-13T16:30:48+00:00","og_image":[{"width":1792,"height":1024,"url":"https:\/\/www.cloudmagazin.com\/wp-content\/uploads\/2026\/07\/soofi-s-souveraen-heisst-nicht-sieger-cover-hero-1.jpg","type":"image\/jpeg"}],"author":"Alec Chizhik","twitter_card":"summary_large_image","twitter_creator":"@cloudmagazin","twitter_site":"@cloudmagazin","twitter_misc":{"Written by":"Alec Chizhik","Est. reading time":"7 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"NewsArticle","@id":"https:\/\/www.cloudmagazin.com\/en\/2026\/07\/13\/soofi-s-sovereign-doesnt-mean-superior\/#article","isPartOf":{"@id":"https:\/\/www.cloudmagazin.com\/en\/2026\/07\/13\/soofi-s-sovereign-doesnt-mean-superior\/"},"author":{"name":"Alec Chizhik","@id":"https:\/\/www.cloudmagazin.com\/en\/#\/schema\/person\/ce38baaa19a580268aedce096597eb3c"},"headline":"Soofi S: Sovereign Doesn\u2019t Mean Superior","datePublished":"2026-07-13T08:59:31+00:00","dateModified":"2026-07-13T16:30:48+00:00","mainEntityOfPage":{"@id":"https:\/\/www.cloudmagazin.com\/en\/2026\/07\/13\/soofi-s-sovereign-doesnt-mean-superior\/"},"wordCount":1486,"publisher":{"@id":"https:\/\/www.cloudmagazin.com\/en\/#organization"},"image":{"@id":"https:\/\/www.cloudmagazin.com\/en\/2026\/07\/13\/soofi-s-sovereign-doesnt-mean-superior\/#primaryimage"},"thumbnailUrl":"https:\/\/www.cloudmagazin.com\/wp-content\/uploads\/2026\/07\/soofi-s-souveraen-heisst-nicht-sieger-cover-hero-1.jpg","articleSection":["Artificial Intelligence","Reboot Germany"],"inLanguage":"en-US"},{"@type":"WebPage","@id":"https:\/\/www.cloudmagazin.com\/en\/2026\/07\/13\/soofi-s-sovereign-doesnt-mean-superior\/","url":"https:\/\/www.cloudmagazin.com\/en\/2026\/07\/13\/soofi-s-sovereign-doesnt-mean-superior\/","name":"Soofi S: Sovereign Doesn\u2019t Mean Superior - cloudmagazin","isPartOf":{"@id":"https:\/\/www.cloudmagazin.com\/en\/#website"},"primaryImageOfPage":{"@id":"https:\/\/www.cloudmagazin.com\/en\/2026\/07\/13\/soofi-s-sovereign-doesnt-mean-superior\/#primaryimage"},"image":{"@id":"https:\/\/www.cloudmagazin.com\/en\/2026\/07\/13\/soofi-s-sovereign-doesnt-mean-superior\/#primaryimage"},"thumbnailUrl":"https:\/\/www.cloudmagazin.com\/wp-content\/uploads\/2026\/07\/soofi-s-souveraen-heisst-nicht-sieger-cover-hero-1.jpg","datePublished":"2026-07-13T08:59:31+00:00","dateModified":"2026-07-13T16:30:48+00:00","description":"Soofi S tops German benchmarks and excels in long-context speed, yet Qwen3.5 dominates reasoning. Explore the full technical performance breakdown.","breadcrumb":{"@id":"https:\/\/www.cloudmagazin.com\/en\/2026\/07\/13\/soofi-s-sovereign-doesnt-mean-superior\/#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/www.cloudmagazin.com\/en\/2026\/07\/13\/soofi-s-sovereign-doesnt-mean-superior\/"]}]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/www.cloudmagazin.com\/en\/2026\/07\/13\/soofi-s-sovereign-doesnt-mean-superior\/#primaryimage","url":"https:\/\/www.cloudmagazin.com\/wp-content\/uploads\/2026\/07\/soofi-s-souveraen-heisst-nicht-sieger-cover-hero-1.jpg","contentUrl":"https:\/\/www.cloudmagazin.com\/wp-content\/uploads\/2026\/07\/soofi-s-souveraen-heisst-nicht-sieger-cover-hero-1.jpg","width":1792,"height":1024,"caption":"Parallele Linien und eine t\u00fcrkisfarbene Kante auf hellem Grund."},{"@type":"BreadcrumbList","@id":"https:\/\/www.cloudmagazin.com\/en\/2026\/07\/13\/soofi-s-sovereign-doesnt-mean-superior\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/www.cloudmagazin.com\/en\/home\/"},{"@type":"ListItem","position":2,"name":"Soofi S: Sovereign Doesn\u2019t Mean Superior"}]},{"@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\/ce38baaa19a580268aedce096597eb3c","name":"Alec Chizhik","image":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/www.cloudmagazin.com\/wp-content\/uploads\/2026\/03\/alec-chizhik.jpg","url":"https:\/\/www.cloudmagazin.com\/wp-content\/uploads\/2026\/03\/alec-chizhik.jpg","contentUrl":"https:\/\/www.cloudmagazin.com\/wp-content\/uploads\/2026\/03\/alec-chizhik.jpg","caption":"Alec Chizhik"},"description":"Alec is the Chief Digital Officer at Evernine and writes about cloud architectures, IT security, and digital operations practices.","sameAs":["https:\/\/www.linkedin.com\/in\/alecchizhik\/"],"url":"https:\/\/www.cloudmagazin.com\/en\/author\/alec\/"}]}},"_links":{"self":[{"href":"https:\/\/www.cloudmagazin.com\/en\/wp-json\/wp\/v2\/posts\/49065","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\/31"}],"replies":[{"embeddable":true,"href":"https:\/\/www.cloudmagazin.com\/en\/wp-json\/wp\/v2\/comments?post=49065"}],"version-history":[{"count":2,"href":"https:\/\/www.cloudmagazin.com\/en\/wp-json\/wp\/v2\/posts\/49065\/revisions"}],"predecessor-version":[{"id":49073,"href":"https:\/\/www.cloudmagazin.com\/en\/wp-json\/wp\/v2\/posts\/49065\/revisions\/49073"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.cloudmagazin.com\/en\/wp-json\/wp\/v2\/media\/49030"}],"wp:attachment":[{"href":"https:\/\/www.cloudmagazin.com\/en\/wp-json\/wp\/v2\/media?parent=49065"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.cloudmagazin.com\/en\/wp-json\/wp\/v2\/categories?post=49065"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.cloudmagazin.com\/en\/wp-json\/wp\/v2\/tags?post=49065"},{"taxonomy":"industry","embeddable":true,"href":"https:\/\/www.cloudmagazin.com\/en\/wp-json\/wp\/v2\/industry?post=49065"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}