
What happened
At Dreamforce, Salesforce and Nvidia unveiled Koa, a domain-specific reasoning model for Salesforce's Agentforce ecosystem. Hosted on Salesforce's private infrastructure, it gives enterprise clients full control over model weights and sensitive data while lowering token inference costs.
Why it matters
This is a notable strategic shift, because many application vendors have spent years distancing themselves from building their own models. The move suggests task-specific models may become central to enterprise AI deployments, as customers seek to cut token costs without sacrificing performance.
What to watch
Whether enterprise clients adopt Koa hinges on whether hosting it on private infrastructure truly delivers lower token costs without hurting performance. Watch how Salesforce's Agentforce ecosystem evolves as it leans on proprietary data and open model approaches.
WHO IT HITSEnterprise AI teams and application vendors are affected, since the move suggests a shift toward domain-specific models and open approaches becoming central to enterprise deployments.
Summaries like this, in your inbox every morning.
Marc Benioff's praise for Nvidia came during Salesforce's annual Dreamforce event, where the two companies unveiled Koa, a domain-specific reasoning model built for Salesforce's Agentforce ecosystem. By hosting Koa on Salesforce's private infrastructure, enterprise clients reportedly maintain full control over model weights and sensitive data while achieving lower token inference costs.
The partnership fits into a broader industry conversation. Guggenheim analyst John Difucci noted that Salesforce's decision to build a proprietary model stands out, since many application vendors have spent years distancing themselves from the model race. Difucci framed the announcement as part of a trend toward task-specific and open-source models becoming key for enterprise deployments, as customers look to save on token costs without sacrificing performance. Stifel analyst J. Parker Lane added that the presence of Jensen Huang, Dario Amodei, and Sam Altman at Dreamforce underscored Salesforce's open approach at the model and UI layers, with the company leaning on its data and context layers as its core moat in the Agentic Era.
Benioff and Salesforce have just wrapped one of the better-received Dreamforce events in recent years, with a keynote and product demos that seemed to quiet fears that model builders would crush software players. Whether Koa's private-infrastructure approach actually delivers the promised cost and performance benefits for enterprise clients, and how Salesforce's Agentforce ecosystem develops around proprietary data, will likely determine how much this partnership reshapes the company's AI position.
For example, today's edition would include:
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. Q&As are published on this page for other readers too.
Doosan committed KRW968 billion to expand high-end copper-clad laminate (CCL) production in South Korea and Ch…

Anthropic Institute said that as of August, 26% of its AI research and development work was almost entirely do…

Abel more than tripled Berkshire's Alphabet stake in the first quarter and added another $17 billion in the se…

Kai-Fu Lee says most CEOs want AI transformation but are 'mostly doing it wrong', and that they must personall…

Atlassian CEO Mike Cannon-Brookes said the most AI-native customers are expanding fastest, and that customers…

Per Betterment's 2026 Retail Investor Survey of 1,000 U.S
