Salesforce and Nvidia announced Koa, described as Salesforce's first CRM reasoning model for its Agentforce platform, built on Nvidia's Nemotron architecture.

Koa was developed by post-training Nvidia Nemotron 3 Super using a proprietary synthetic dataset drawn from what Salesforce characterizes as nearly 27 years of CRM deployment knowledge. No customer data was used in training. According to Salesforce's internal CRM Benchmark, Koa matches or exceeds leading model performance on CRM-related tasks with three times fewer errors.

The model is designed to handle multi-step enterprise workflows across more than 14 industries, including manufacturing, financial services, healthcare, and travel. Salesforce controls the model weights and runs inference within its own infrastructure.

Koa is currently deployed internally at Salesforce, including an agent in Slack, and is entering customer pilots with 1-800Accountant, Baxter Credit Union, Engine, Formula 1, UChicago Medicine, and Xero. General availability is expected in Winter 2026 in U.S. regions.

The two companies are also integrating Nvidia models and accelerated computing into Missionforce, Salesforce's product for government and regulated organizations. This is intended to allow those customers to train and deploy models in private clouds, air-gapped networks, and other secure environments. Missionforce Operations is generally available now in U.S. regions, with post-trained Nvidia models expected to reach select customers in October 2026.

"With Koa, the knowledge is put inside the model itself," said Marc Benioff, Chair and CEO of Salesforce. "We trained a reasoning engine that understands the structure of a deal, the lifecycle of a service case, and the workflows that vary across industries."

Jensen Huang, founder and CEO of Nvidia, said: "NVIDIA Nemotron open models give Salesforce the foundation to turn decades of enterprise expertise into specialized AI with Koa, creating a CRM model that can reason and securely take action."