1d:I[70616,["/_next/static/chunks/0kcf5dt6yxdox.js","/_next/static/chunks/1su5n1uv1ti4s.js","/_next/static/chunks/0gvke1wv6udfg.js","/_next/static/chunks/1dfrle3jgmlzh.js","/_next/static/chunks/162lgvduyr-fa.js","/_next/static/chunks/2gcnnopam8e9w.js","/_next/static/chunks/2g60iab5a1pkg.js"],"Breadcrumb"] 1e:I[90762,["/_next/static/chunks/0kcf5dt6yxdox.js","/_next/static/chunks/1su5n1uv1ti4s.js","/_next/static/chunks/0gvke1wv6udfg.js","/_next/static/chunks/1dfrle3jgmlzh.js","/_next/static/chunks/162lgvduyr-fa.js","/_next/static/chunks/2gcnnopam8e9w.js","/_next/static/chunks/2g60iab5a1pkg.js"],"ArticleGrid"] 1f:T190f,

The most interesting figure in Nvidia and Palantir's September 10 supply-chain announcement is not a dollar amount or a customer count. It is an accuracy gap: on the job of deciding which scarce components go where, a 30-billion-parameter model that had been taught Nvidia's own past decisions scored 86.7%, while a 550-billion-parameter general-purpose sibling managed 55.5%.

Key takeaways

Post-trained Nemotron 3.5 Lightning (30B) reached 86.7% on Nvidia's materials-allocation task; Nemotron 3 Ultra (550B) reached 55.5% on the same work.
Nvidia's engineers explicitly capped the claim, reporting that production-risk forecasting stayed unreliable even after the same fine-tuning run.
Palantir customers can reproduce the pattern on private data and keep weights, data, and inference inside their own infrastructure.

Why the parameter count lost

Roughly eighteen-to-one is the size ratio between the two models, which makes the result read like an argument against scale. It is better read as an argument about scope. Nvidia solutions architects Nell Barber, Rana Haber, and Aastha Jhunjhunwala attached an unusually candid caveat to their own benchmark: the advantage is confined to the one domain the model was post-trained on, and forecasting which production lines were about to run into trouble remained hard regardless. In other words, fine-tuning moved a decision task and left an adjacent prediction task roughly where it was.

That asymmetry is the practical finding for anyone sizing a deployment. Allocation is a bounded problem with a long internal record of correct answers, which is exactly the shape that rewards specialization. Forecasting is not.

How the stack is wired

Nemotron is the smallest component in the system rather than its engine. Palantir's Foundry and Artificial Intelligence Platform supply the substrate, and the company's Ontology keeps a live structural map of parts, plants, spare capacity, and commitments already promised to customers. Nvidia's cuOpt solver does the combinatorial work of splitting a short supply across competing demands.

The language model's contribution is judgment at the end of that chain, reading the surrounding situation and proposing an action for a human planner to accept or override. Both companies are selling the assembly as a single command surface for supply teams, with materials allocation as the opening use case rather than the whole product.

Why Nvidia tested on itself

A $5.4 trillion market capitalization buys an unusually punishing test environment. Thousands of suppliers and a worldwide manufacturing footprint feed racks that Nvidia counts in the millions of parts, roughly 1.3 million for a single Vera Rubin unit, and the failure mode is brutally simple: one absent component freezes assembly while every part that did show up depreciates in a warehouse.

Jensen Huang, Nvidia's founder and chief executive, pitched that coordination burden as the real bottleneck in AI buildouts, describing supply chains as the operating system of the physical economy and AI factories as among the most complex systems ever assembled. Palantir chief executive Alex Karp, predictably, preferred to emphasize the difficulty of the customer he had landed, characterizing Nvidia's supply chain as arguably the world's most intricate.

What a buyer can actually do with this

Neither the partnership nor the sovereignty framing started here. Nvidia and Palantir tied their models and software together last October, broadened that into controlled-environment deployments in June, and have now produced the reference build with a published score on it. The sequence matters commercially: the June announcement was a capability, and this one is a case study with a number.

Reproducing it means training Nemotron on proprietary operational history through Foundry and AIP, then hosting the result on-premises, in a colocation facility, or with a cloud provider, depending on how tightly the data has to be held. Agriculture, manufacturing, pharmaceuticals, retail, technology, and government are the named targets, and the stack is slated for a turn at Palantir's AIPCon 11.

None of that portability would work without Nvidia publishing weights and, for much of the Nemotron family, training recipes as well. The company has spent a year lobbying for frontier open-weight models on exactly these grounds, then spent $12.9 billion acquiring Hugging Face in early September. A supply-chain deployment where the small open model wins is, conveniently, the best possible advertisement for that position. The primary source is the companies' joint newsroom release; The New Stack's report adds the benchmark detail.

FAQ

Is Nemotron 3.5 Lightning open source?

Open weights is the accurate term. Nvidia distributes the weights, and for much of the family the training data and recipes too, which is what makes private fine-tuning possible. The governing terms are Nvidia's own model license rather than an OSI-approved one.

Does this mean companies should stop using large models?

No. The 86.7% figure covers one narrow allocation task with abundant labeled history, and Nvidia's own write-up says the gains do not generalize. A forecasting task inside the same project resisted the same treatment.

What does sovereign mean in this context?

It refers to control rather than geography: the customer's proprietary data, the model weights, and the inference all stay inside infrastructure the customer governs, whether that is an on-premises cluster or a dedicated colocation footprint.

a:["$","div",null,{"className":"mx-auto max-w-7xl px-4 py-8 sm:px-6","children":[["$","script",null,{"type":"application/ld+json","dangerouslySetInnerHTML":{"__html":"{\"@context\":\"https://schema.org\",\"@type\":\"BreadcrumbList\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https://ainewsway.com\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"#nemotron\",\"item\":\"https://ainewsway.com/tag/nemotron\"}]}"}}],["$","script",null,{"type":"application/ld+json","dangerouslySetInnerHTML":{"__html":"{\"@context\":\"https://schema.org\",\"@type\":\"ItemList\",\"@id\":\"https://ainewsway.com/tag/nemotron#itemlist\",\"name\":\"#nemotron Articles\",\"numberOfItems\":1,\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"url\":\"https://ainewsway.com/articles/a-30b-model-outscored-a-550b-one-inside-nvidias-own-factory-planning-mtzcqmyq\",\"name\":\"A 30B Model Outscored a 550B One Inside Nvidia's Own Factory Planning\"}]}"}}],["$","$L1d",null,{"items":[{"label":"# nemotron"}]}],["$","h1",null,{"className":"mb-8 text-3xl font-bold text-zinc-900 dark:text-zinc-100","children":["#","nemotron"]}],[["$","$L1e",null,{"articles":[{"id":"35f1d07b-12e7-4f99-acbc-c2d78b692987","slug":"a-30b-model-outscored-a-550b-one-inside-nvidias-own-factory-planning-mtzcqmyq","title":"A 30B Model Outscored a 550B One Inside Nvidia's Own Factory Planning","subtitle":"Nvidia and Palantir published an accuracy gap that argues narrow post-training now buys more than raw parameter count on bounded operational work","content":"$1f","excerpt":"Nvidia and Palantir published an 86.7% versus 55.5% accuracy gap favoring a 30B fine-tuned model over a 550B general one on Nvidia's own materials allocation.","coverImage":"https://cdn.ainewsway.com/uploads/2026/09/421f1375-b8b8-4d66-8f5f-c5fe0dcf924b.webp","coverImageCaption":"Nvidia's Santa Clara headquarters, where the supply-chain operations team's historical decisions became the training data for the Nemotron model now recommending allocation calls.","authorId":"6773f67d-d7fa-4b19-8f38-384f10ac761b","status":"PUBLISHED","publishedAt":"2026-09-13T05:07:58.286Z","viewCount":0,"likeCount":0,"readingTime":5,"isFeatured":false,"isBreakingNews":false,"metaTitle":"30B Nemotron Outscores 550B Model on Nvidia Supply Chain","metaDescription":"A fine-tuned 30B Nemotron hit 86.7% on Nvidia's materials allocation task versus 55.5% for a 550B model, in Nvidia and Palantir's sovereign AI deployment.","metaKeywords":["Nvidia Palantir","Nemotron 3.5 Lightning","sovereign AI","supply chain AI","open weight models","Palantir Foundry","fine-tuning","cuOpt"],"ogImage":null,"aiSummary":"Nvidia and Palantir reported on September 10, 2026 that a 30-billion-parameter Nemotron 3.5 Lightning model, post-trained on Nvidia's historical supply decisions, scored 86.7% on materials allocation versus 55.5% for a 550-billion-parameter general model. Nvidia's engineers cautioned the gain does not generalize, as production-risk forecasting stayed unreliable. Palantir customers can rebuild the system on private data and keep weights and inference on their own infrastructure.","aiTags":[],"sentiment":null,"articleType":"NEWS","videoEmbed":null,"videoType":null,"seoScore":85,"isAutoGenerated":true,"isBackfill":false,"pipelineStage":null,"copyrightScore":null,"similarityScore":null,"matchedSourceId":null,"createdAt":"2026-09-13T05:06:52.228Z","updatedAt":"2026-09-13T09:06:45.485Z","contentUpdatedAt":"2026-09-13T09:06:45.484Z","deletedAt":null,"author":{"id":"6773f67d-d7fa-4b19-8f38-384f10ac761b","name":"Seung Jung","username":"Seung Jung","avatar":null},"categories":[{"id":"8126f4b3-ce4a-464f-bc70-7c3c39d5540b","name":"Tech & Business","slug":"tech-business"}],"tags":[{"id":"51075b6e-7339-4ca5-a00d-1b068653c310","name":"NVIDIA","slug":"nvidia"},{"id":"d601177f-b52b-4344-9264-8fa680370c26","name":"Palantir","slug":"palantir"},{"id":"776049c1-20c7-4b1a-81cd-d40058dd8281","name":"Nemotron","slug":"nemotron"},{"id":"c25252f9-3559-4d8a-a091-cf0c2f14468b","name":"Sovereign AI","slug":"sovereign-ai"},{"id":"312dd740-62ff-4e74-b41b-149482a8caa0","name":"Supply Chain AI","slug":"supply-chain-ai"},{"id":"0aad1028-055f-4322-b7ce-a7ce5b46a44c","name":"Open Weights","slug":"open-weights"}],"translations":[{"id":"33eebbdb-3045-4cf4-aa00-cb849aa8936b","language":"ko","title":"엔비디아 자체 공장 계획에서 30B 모델이 550B 모델을 앞질렀다","subtitle":"엔비디아와 팔란티어가 공개한 정확도 격차는 한정된 운영 업무에서 파라미터 규모보다 좁은 사후학습이 더 큰 값을 한다는 주장을 뒷받침한다","excerpt":"엔비디아와 팔란티어가 자사 자재 배분 업무에서 미세조정한 30B 모델이 86.7%, 범용 550B 모델이 55.5%를 기록한 정확도 격차를 공개했다.","metaTitle":"30B 네모트론, 엔비디아 공급망에서 550B 모델 제쳤다","metaDescription":"미세조정한 30B 네모트론이 엔비디아 자재 배분 업무에서 86.7%를 기록했다. 같은 업무에서 550B 모델은 55.5%에 그쳤다. 엔비디아와 팔란티어의 소버린 AI 구축 사례다."},{"id":"e0535503-9e49-4a1e-929f-834c04e9c42f","language":"ja","title":"エヌビディア自社の工場計画で、30Bモデルが550Bモデルを上回った","subtitle":"エヌビディアとパランティアが公表した精度差は、範囲の限られた業務ではパラメータ数より狭い事後学習のほうが効くという主張を裏づける","excerpt":"エヌビディアとパランティアが、自社の資材配分業務でファインチューニングした30Bモデルが86.7%、汎用の550Bモデルが55.5%という精度差を公表した。","metaTitle":"30B Nemotronがエヌビディア供給網で550Bモデルを上回る","metaDescription":"ファインチューニングした30B Nemotronがエヌビディアの資材配分タスクで86.7%を記録。同じ業務で550Bモデルは55.5%だった。両社の主権AI展開の事例。"},{"id":"68e1b3b7-14c0-4c8f-ae4e-2be15ad0127a","language":"zh","title":"在英伟达自家的工厂调度里，300亿参数模型跑赢了5500亿","subtitle":"英伟达与Palantir公开的精度差距表明，在边界清晰的运营任务上，窄领域后训练比堆参数更划算","excerpt":"英伟达与Palantir公布，在自家物料分配任务上，微调后的30B模型拿到86.7%，通用的550B模型只有55.5%。","metaTitle":"30B Nemotron在英伟达供应链任务上跑赢550B模型","metaDescription":"微调后的30B Nemotron在英伟达物料分配任务上取得86.7%，同一任务的550B模型为55.5%，出自两家公司的主权AI部署案例。"},{"id":"cc9bad4d-50ad-4e02-8cec-e38863a1b300","language":"es","title":"Un modelo de 30B superó a uno de 550B en la propia planificación fabril de Nvidia","subtitle":"La brecha de precisión que publicaron Nvidia y Palantir sugiere que, en trabajo operativo acotado, el postentrenamiento estrecho rinde más que el recuento de parámetros","excerpt":"Nvidia y Palantir publicaron una brecha de 86,7% frente a 55,5% a favor de un modelo de 30B ajustado sobre uno general de 550B en su propia asignación de materiales.","metaTitle":"Nemotron 30B supera a un modelo de 550B en Nvidia","metaDescription":"Un Nemotron de 30B ajustado alcanzó 86,7% en la asignación de materiales de Nvidia frente al 55,5% de un modelo de 550B, en el despliegue de IA soberana con Palantir."}],"celebrities":[{"id":"95f8f574-a1e4-4a77-8783-1419c1421b78","nameKo":"엔비디아","nameEn":"NVIDIA","slug":"nvidia","profileImage":null},{"id":"bced8c1e-cbb8-431b-90be-ebe05c8932fe","nameKo":"팔란티어","nameEn":"Palantir Technologies","slug":"palantir-technologies","profileImage":null},{"id":"88699bff-488e-4c7b-933f-c4d7f238d84b","nameKo":"젠슨 황","nameEn":"Jensen Huang","slug":"jensen-huang","profileImage":"https://ui-avatars.com/api/?name=Jensen+Huang&size=400&background=EF4444&color=fff&bold=true"}]}]}],"$L20"]]}]

#nemotron

엔비디아 자체 공장 계획에서 30B 모델이 550B 모델을 앞질렀다

뉴스레터 구독

#nemotron

엔비디아 자체 공장 계획에서 30B 모델이 550B 모델을 앞질렀다