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:T17a6,

Salesforce and Nvidia used the opening of Dreamforce in San Francisco on September 15 to announce Koa, the CRM vendor's first reasoning model, post-trained from Nvidia's open-weight Nemotron 3 Super and run entirely inside Salesforce's own infrastructure. It is the first time Salesforce has stopped renting reasoning from a frontier lab and built it in-house.

Key takeaways

Koa was post-trained from Nvidia Nemotron 3 Super on a synthetic corpus modeled on 27 years of Salesforce CRM deployments, with no customer data in the training set.
On Salesforce's internal CRM benchmark, Koa matches or exceeds leading models on CRM actions while producing three times fewer errors.
Koa is in pilot with 1-800Accountant, Baxter Credit Union, Engine, Formula 1, UChicago Medicine and Xero, with general availability expected in winter 2026 across U.S. regions.

Why Salesforce stopped renting reasoning

Agentforce, the platform where Salesforce customers assemble AI agents for rote work like case routing and appointment scheduling, already shipped a family of small, task-specific models. Reasoning was the gap. Whenever an agent hit a long-running or multi-step problem, Agentforce's model gateway handed the prompt off to Claude or ChatGPT.

Jayesh Govindarajan, Salesforce's EVP of AI, told TechCrunch that the company had wanted to train its own enterprise-grade reasoning model for some time but lacked a usable starting point. What Nemotron supplied, in his telling, was a base model that was simultaneously American-hosted, state of the art, and transparent about what went into it — a contrast he drew explicitly against Alibaba's Qwen, whose training data he said is unknown.

How Koa was trained without customer data

None of Salesforce's customer records were used. Instead the company generated synthetic scenarios spanning more than 14 industries, including manufacturing, financial services, healthcare and travel. Each scenario paired a persona — an irate caller into a support line, a rep trying to close a deal — with a task, then mapped the exact sequence of actions and tool calls an agent would need to finish it.

The post-training itself combined supervised fine-tuning with reinforcement learning using Group Relative Policy Optimization, run on Nvidia's NeMo RL, NeMo Gym and NeMo AutoModel stack. Salesforce says the narrow task set is what produced the error gap: the model learned not just to answer correctly but to sequence the right actions toward a goal.

What changes for enterprise buyers

The pitch is control and cost rather than raw capability. Salesforce holds the weights and performs both post-training and inference inside its own trust boundary, so no customer data crosses out during a request. Because Koa is tuned for a bounded set of CRM actions, it also burns fewer tokens than routing the same work to a general-purpose frontier model.

Kari Ann Briski, Nvidia's VP of generative AI software for enterprise, framed Nemotron's inference architecture as delivering a combination enterprises need at once — sovereign deployment, fast time to first token, and efficient reasoning economics. Marc Benioff put it in blunter terms, arguing that Salesforce's most valuable asset was never the platform but the accumulated knowledge of how enterprise business runs, and that Koa moves that knowledge inside the model.

The same collaboration extends to Missionforce, Salesforce's public-sector arm, where Nemotron-derived models are being deployed into private clouds, classified networks and fully air-gapped systems for government procurement, supplier management and logistics workflows.

Outlook

Salesforce is not walking away from the labs. Days earlier it announced Claudeforce with Anthropic, letting companies use Claude as their interface while records stay in Salesforce's system. Koa is positioned as another option in the gateway, not a replacement for it. But the direction is unmistakable, and it echoes a broader pattern in which Nvidia's open models keep turning up underneath enterprise agent stacks — as in the 30B Nemotron-and-router pairing the company shipped to cut agent costs. Koa reaches general availability in winter 2026; Missionforce Operations is generally available now, with post-trained Nvidia models opening to select customers in October.

FAQ

Is Koa open source?

No. Koa is built on Nvidia's open-weight Nemotron 3 Super, but the post-trained result is not released. Salesforce controls the weights and runs both training and inference on its own infrastructure, and customers reach the model through Agentforce rather than downloading it.

Was any Salesforce customer data used to train Koa?

Salesforce says no. The training corpus was built entirely from synthetic scenarios designed to mirror how CRM work actually unfolds across industries. The company presents this as the reason Koa cannot leak one customer's data to another.

When can companies use Koa?

Select pilot customers have access inside Agentforce now, including 1-800Accountant, Baxter Credit Union, Engine, Formula 1, UChicago Medicine and Xero. General availability is expected in winter 2026 in U.S. regions.

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\":\"#reasoning-model\",\"item\":\"https://ainewsway.com/tag/reasoning-model\"}]}"}}],["$","script",null,{"type":"application/ld+json","dangerouslySetInnerHTML":{"__html":"{\"@context\":\"https://schema.org\",\"@type\":\"ItemList\",\"@id\":\"https://ainewsway.com/tag/reasoning-model#itemlist\",\"name\":\"#reasoning-model Articles\",\"numberOfItems\":1,\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"url\":\"https://ainewsway.com/articles/salesforce-trained-its-own-reasoning-model-on-nvidia-nemotron-and-kept-the-weights-mu3vkmfy\",\"name\":\"Salesforce Trained Its Own Reasoning Model on Nvidia Nemotron — and Kept the Weights\"}]}"}}],["$","$L1d",null,{"items":[{"label":"# reasoning-model"}]}],["$","h1",null,{"className":"mb-8 text-3xl font-bold text-zinc-900 dark:text-zinc-100","children":["#","reasoning-model"]}],[["$","$L1e",null,{"articles":[{"id":"e5a32da5-9e9c-4a8f-b561-91e09b0015f8","slug":"salesforce-trained-its-own-reasoning-model-on-nvidia-nemotron-and-kept-the-weights-mu3vkmfy","title":"Salesforce Trained Its Own Reasoning Model on Nvidia Nemotron — and Kept the Weights","subtitle":"Koa matches leading models on CRM actions with three times fewer errors, Salesforce says","content":"$1f","excerpt":"Salesforce unveiled Koa at Dreamforce: a CRM reasoning model post-trained from Nvidia Nemotron 3 Super, trained on synthetic data and hosted in-house.","coverImage":"https://cdn.ainewsway.com/uploads/2026/09/ac62b59d-948b-4d4b-89ac-2d30004d17da.webp","coverImageCaption":"Salesforce Tower in San Francisco, where the company opened Dreamforce 2026 by announcing Koa, its first in-house CRM reasoning model","authorId":"6773f67d-d7fa-4b19-8f38-384f10ac761b","status":"PUBLISHED","publishedAt":"2026-09-16T09:06:59.163Z","viewCount":0,"likeCount":0,"readingTime":4,"isFeatured":false,"isBreakingNews":false,"metaTitle":"Salesforce Koa: CRM Reasoning Model on Nvidia Nemotron","metaDescription":"Salesforce Koa, post-trained from Nvidia Nemotron 3 Super on synthetic CRM data, matches leading models with 3x fewer errors. GA expected winter 2026.","metaKeywords":["Salesforce Koa","Nvidia Nemotron","CRM reasoning model","Agentforce","enterprise AI agents","open-weight model","Dreamforce 2026"],"ogImage":null,"aiSummary":"Salesforce announced Koa on September 15 at Dreamforce, its first CRM reasoning model, post-trained from Nvidia's open-weight Nemotron 3 Super on synthetic data modeled on 27 years of CRM deployments. On Salesforce's internal benchmark it matches or exceeds leading models on CRM actions with three times fewer errors. The move signals enterprises building specialized reasoning in-house rather than routing every complex task to frontier labs. General availability is expected in winter 2026.","aiTags":[],"sentiment":null,"articleType":"NEWS","videoEmbed":null,"videoType":null,"seoScore":85,"isAutoGenerated":true,"isBackfill":false,"pipelineStage":null,"copyrightScore":null,"similarityScore":null,"matchedSourceId":null,"createdAt":"2026-09-16T09:05:09.023Z","updatedAt":"2026-09-16T12:12:28.772Z","contentUpdatedAt":"2026-09-16T12:12:28.771Z","deletedAt":null,"author":{"id":"6773f67d-d7fa-4b19-8f38-384f10ac761b","name":"Seung Jung","username":"Seung Jung","avatar":null},"categories":[{"id":"e2b5fffc-de32-42aa-86fe-c75f219d6b4b","name":"AI & Machine Learning","slug":"ai-ml"}],"tags":[{"id":"45731b2f-f426-4e83-bb26-c81e698c6ea4","name":"Salesforce","slug":"salesforce"},{"id":"51075b6e-7339-4ca5-a00d-1b068653c310","name":"NVIDIA","slug":"nvidia"},{"id":"776049c1-20c7-4b1a-81cd-d40058dd8281","name":"Nemotron","slug":"nemotron"},{"id":"97ca3168-2722-484b-8cb8-970bbe0f08dd","name":"Agentforce","slug":"agentforce"},{"id":"c8881e00-ac7f-4ff1-9380-00a5987e7b65","name":"Reasoning Model","slug":"reasoning-model"},{"id":"582ce609-bf37-43c0-9333-d510071c4209","name":"Enterprise AI","slug":"enterprise-ai"}],"translations":[{"id":"69552f56-161b-48d4-a596-b01396d7c163","language":"es","title":"Salesforce entrenó su propio modelo de razonamiento sobre Nvidia Nemotron y se quedó con los pesos","subtitle":"Koa iguala a los modelos líderes en acciones de CRM con tres veces menos errores, según Salesforce","excerpt":"Salesforce presentó Koa en Dreamforce: un modelo de razonamiento para CRM entrenado a partir de Nvidia Nemotron 3 Super con datos sintéticos y alojado en su propia infraestructura.","metaTitle":"Salesforce Koa: razonamiento CRM sobre Nvidia Nemotron","metaDescription":"Koa, entrenado sobre Nvidia Nemotron 3 Super con datos sintéticos de CRM, iguala a los líderes con tres veces menos errores. Llega en invierno de 2026."},{"id":"47a81e61-f0e4-487d-9717-7d796067e2bd","language":"ja","title":"Salesforce、Nvidia Nemotronで自社推論モデルを開発し重みも手放さず","subtitle":"KoaはCRM処理で主要モデルに並びながらエラーは3分の1だとSalesforceは説明する","excerpt":"SalesforceがDreamforceでKoaを発表した。Nvidia Nemotron 3 Superを合成データで追加学習したCRM推論モデルで、自社インフラ上で稼働する。","metaTitle":"Salesforce Koa、Nemotron由来のCRM推論モデル","metaDescription":"Nvidia Nemotron 3 SuperをCRM合成データで追加学習したSalesforce Koaは、主要モデル並みの性能でエラー3分の1。正式提供は2026年冬の見込み。"},{"id":"b560f9f7-9e48-4b68-ad3a-a65406929643","language":"ko","title":"Salesforce, Nvidia Nemotron으로 자체 추론 모델 만들고 가중치도 직접 쥐었다","subtitle":"Koa는 CRM 작업에서 주요 모델과 대등하면서 오류는 3분의 1 수준이라고 Salesforce는 밝혔다","excerpt":"Salesforce가 드림포스에서 Koa를 공개했다. Nvidia Nemotron 3 Super를 합성 데이터로 후속 학습한 CRM 추론 모델로, 자체 인프라에서 돌아간다.","metaTitle":"Salesforce Koa, Nvidia Nemotron 기반 CRM 추론 모델","metaDescription":"Nvidia Nemotron 3 Super를 CRM 합성 데이터로 후속 학습한 Salesforce Koa가 주요 모델과 대등하면서 오류는 3분의 1이다. 정식 출시는 2026년 겨울."},{"id":"147477fb-3014-4ea7-b44a-ffe2b1576fb0","language":"zh","title":"Salesforce 用 Nvidia Nemotron 訓出自家推理模型，權重也自己留著","subtitle":"Salesforce 表示，Koa 在 CRM 操作上追平主流模型，錯誤卻只有三分之一","excerpt":"Salesforce 在 Dreamforce 發表 Koa：以 Nvidia Nemotron 3 Super 後訓練而成的 CRM 推理模型，用合成資料訓練，並在自家基礎設施上託管。","metaTitle":"Salesforce Koa：Nemotron 打底的 CRM 推理模型","metaDescription":"Salesforce Koa 以 Nvidia Nemotron 3 Super 搭配 CRM 合成資料後訓練，表現追平主流模型且錯誤少三分之二，預計 2026 年冬季正式上線。"}],"celebrities":[{"id":"2f86403d-9b81-4d10-8461-4a69ea91f657","nameKo":"세일즈포스","nameEn":"Salesforce","slug":"salesforce","profileImage":null},{"id":"95f8f574-a1e4-4a77-8783-1419c1421b78","nameKo":"엔비디아","nameEn":"NVIDIA","slug":"nvidia","profileImage":null},{"id":"7e36dd51-a78e-40fc-ab21-fa676433db90","nameKo":"마크 베니오프","nameEn":"Marc Benioff","slug":"marc-benioff","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"]]}]

#reasoning-model

Salesforce entrenó su propio modelo de razonamiento sobre Nvidia Nemotron y se quedó con los pesos

Suscríbete al Boletín

#reasoning-model

Salesforce entrenó su propio modelo de razonamiento sobre Nvidia Nemotron y se quedó con los pesos