GoKawiilSalesforce debuts Koa, a reasoning AI model built on Nvidia's Nemotron for Agentforce
Salesforce unveiled Koa, its first reasoning model, at Dreamforce this week. Built on Nvidia's open-weight Nemotron model and post-trained for sales, marketing, and customer-support tasks, Koa gives Salesforce customers an alternative to routing complex agent requests through frontier models like Claude or ChatGPT.
GoKawiil's interpretation of the reporting above, not reported fact.
Koa signals that enterprise AI needs are pulling away from what frontier labs like OpenAI and Anthropic offer, favoring task-specific, cost-efficient models over general-purpose reasoning systems trained on customer data. By keeping data in-house and reducing token usage, Salesforce is positioning itself to cut reliance on costly third-party model providers while tightening control over security and compliance.
- Koa is Salesforce's first reasoning model, built on Nvidia's Nemotron architecture.
- It targets sales, marketing, and customer-support tasks without ingesting customer data.
- Koa reduces AI spending by using fewer tokens and can be routed automatically via Agentforce's AI gateway.
Source: techcrunch.com — Julie Bort, 2026-09-15
Published there as: “Salesforce and Nvidia’s new reasoning model is everything the AI labs should fear”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.