--- title: Salesforce and Nvidia’s new reasoning model is... description: Salesforce unveiled Koa, its first reasoning model, built on Nvidia's open-weight Nemotron model and post-trained with synthetic data to handle sales,... canonical: https://daily.dev/posts/salesforce-and-nvidia-s-new-reasoning-model-is-everything-the-ai-labs-should-fear-fpg7clt9f twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Salesforce and Nvidia’s new reasoning model is everything the AI labs should fear | daily.dev og:description: Salesforce unveiled Koa, its first reasoning model, built on Nvidia's open-weight Nemotron model and post-trained with synthetic data to handle sales,... og:url: https://daily.dev/posts/salesforce-and-nvidia-s-new-reasoning-model-is-everything-the-ai-labs-should-fear-fpg7clt9f og:image: https://api.daily.dev/og/posts/FpG7clt9F.png og:image:alt: Salesforce and Nvidia’s new reasoning model is everything the AI labs should fear og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Salesforce and Nvidia’s new reasoning model is everything the AI labs should fear **[TechCrunch](https://daily.dev/sources/tc)** · 4 min read · 0 upvotes · 0 comments ## Summary Salesforce unveiled Koa, its first reasoning model, built on Nvidia's open-weight Nemotron model and post-trained with synthetic data to handle sales, marketing, and customer-support tasks. Unlike frontier models from Anthropic or OpenAI, Koa avoids ingesting real customer data, uses fewer tokens per task, and can be routed through Salesforce's Agentforce AI gateway. Salesforce says it chose Nemotron because it needed a sovereign, state-of-the-art, pre-trained base model with clear data provenance, contrasting it with alternatives like Alibaba's Qwen. Salesforce is not abandoning frontier labs entirely, having also launched a ClaudeForce partnership with Anthropic that keeps customer data inside Salesforce's own infrastructure while using Claude as an interface. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What is Salesforce's Koa model and what is it built on? Koa is Salesforce's first reasoning model, built on Nvidia's open-weight Nemotron model and post-trained by Salesforce and Nvidia to handle sales, marketing, and customer-support tasks. It was trained using synthetic data simulating customer service and sales scenarios rather than actual customer data, so it cannot leak real customer information, and it uses fewer tokens than routing the same tasks to frontier models. _Teams weighing task-specific versus frontier reasoning models can track releases like Koa on daily.dev._ ### Why did Salesforce choose Nvidia's Nemotron instead of another open-weight model like Qwen for its reasoning model? Salesforce chose Nemotron because it was the first sovereign American pre-trained base model that was both state-of-the-art and had clear data provenance, according to Salesforce AI EVP Jayesh Govindarajan. He noted that alternatives like Alibaba's Qwen lack transparency about what data they were trained on, which mattered for an enterprise-grade model handling sensitive business tasks. _Developers picking open-weight base models for enterprise use can follow this reasoning on daily.dev._ ### How does Koa fit alongside Claude and ChatGPT in Salesforce's Agentforce platform? Koa is offered as an alternative to frontier models like Claude and ChatGPT within Agentforce's AI gateway, which routes prompts to whichever model suits the task. Previously, long-running or multi-step reasoning tasks were sent to frontier providers; now Koa can handle many of those directly, though Salesforce also launched ClaudeForce, a partnership letting customers use Claude as an interface while data stays within Salesforce's infrastructure. _Anyone architecting multi-model AI routing can follow these platform choices on daily.dev._ ## Similar posts on daily.dev - [Salesforce is crowdsourcing its AI roadmap — with customers](https://daily.dev/posts/salesforce-is-crowdsourcing-its-ai-roadmap-with-customers-busiahsw3) · TechCrunch · 2 upvotes · 0 comments --- Tags: [#nvidia](https://daily.dev/tags/nvidia), [#salesforce](https://daily.dev/tags/salesforce) [View this post on daily.dev](https://daily.dev/posts/salesforce-and-nvidia-s-new-reasoning-model-is-everything-the-ai-labs-should-fear-fpg7clt9f) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Salesforce and Nvidia’s new reasoning model is everything the AI labs should fear","url":"https://daily.dev/posts/salesforce-and-nvidia-s-new-reasoning-model-is-everything-the-ai-labs-should-fear-fpg7clt9f","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/salesforce-and-nvidia-s-new-reasoning-model-is-everything-the-ai-labs-should-fear-fpg7clt9f"},"datePublished":"2026-09-15T12:03:58.084Z","dateModified":"2026-09-16T16:41:37.975Z","description":"Salesforce unveiled Koa, its first reasoning model, built on Nvidia's open-weight Nemotron model and post-trained with synthetic data to handle sales,...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/dd731a1fc18362de23b7a4e1b6e2f884?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/dd731a1fc18362de23b7a4e1b6e2f884?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"TechCrunch","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"TechCrunch","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/tc","url":"https://daily.dev/sources/tc"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/salesforce-and-nvidia-s-new-reasoning-model-is-everything-the-ai-labs-should-fear-fpg7clt9f","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"nvidia,salesforce","timeRequired":"PT4M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"TechCrunch","item":"https://daily.dev/sources/tc"},{"@type":"ListItem","position":3,"name":"Salesforce and Nvidia’s new reasoning model is everything the AI labs should fear"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/salesforce-and-nvidia-s-new-reasoning-model-is-everything-the-ai-labs-should-fear-fpg7clt9f#faq","mainEntity":[{"@type":"Question","name":"What is Salesforce's Koa model and what is it built on?","acceptedAnswer":{"@type":"Answer","text":"Koa is Salesforce's first reasoning model, built on Nvidia's open-weight Nemotron model and post-trained by Salesforce and Nvidia to handle sales, marketing, and customer-support tasks. It was trained using synthetic data simulating customer service and sales scenarios rather than actual customer data, so it cannot leak real customer information, and it uses fewer tokens than routing the same tasks to frontier models. Teams weighing task-specific versus frontier reasoning models can track releases like Koa on daily.dev."}},{"@type":"Question","name":"Why did Salesforce choose Nvidia's Nemotron instead of another open-weight model like Qwen for its reasoning model?","acceptedAnswer":{"@type":"Answer","text":"Salesforce chose Nemotron because it was the first sovereign American pre-trained base model that was both state-of-the-art and had clear data provenance, according to Salesforce AI EVP Jayesh Govindarajan. He noted that alternatives like Alibaba's Qwen lack transparency about what data they were trained on, which mattered for an enterprise-grade model handling sensitive business tasks. Developers picking open-weight base models for enterprise use can follow this reasoning on daily.dev."}},{"@type":"Question","name":"How does Koa fit alongside Claude and ChatGPT in Salesforce's Agentforce platform?","acceptedAnswer":{"@type":"Answer","text":"Koa is offered as an alternative to frontier models like Claude and ChatGPT within Agentforce's AI gateway, which routes prompts to whichever model suits the task. Previously, long-running or multi-step reasoning tasks were sent to frontier providers; now Koa can handle many of those directly, though Salesforce also launched ClaudeForce, a partnership letting customers use Claude as an interface while data stays within Salesforce's infrastructure. Anyone architecting multi-model AI routing can follow these platform choices on daily.dev."}}]} ```