Cassava adds Nvidia Nemotron open models to its AI system

Cassava Technologies, a technology company, added Nvidia Nemotron open models to its AI Factory. This is its set of computers for AI. Companies use the models through APIs (software parts that let other software use a system) on the Cassava AI Multi-Model Exchange. They do not make model servers. Cassava gave no test results and no prices.

How it works

Cassava operates the Nemotron models in its system. Its Token Optimiser finds data that the system read before and uses it again. Cassava tells that this works with its models and with models from other companies.

The proof

Cassava gave no numbers for cost, speed or token savings. It tells that Nemotron is good at reasoning and tool use. Developers can tune the open models for their agents. An agent is software that does tasks with no person to help.

What is not known

Cassava did not tell the price of the system, which customers use it or how much the Token Optimiser cuts. It uses Nvidia OpenShell for protection and control of agent work. It gave no test results for OpenShell.

Insight: Local hosts compete to give Nvidia open models

In October 2026, Trend AI put Nemotron in customer data centers. Nvidia tests showed that its Rerank model cut agent cost by about 14%. Cassava gives Nemotron with a token tool. This reads as a race between hosts that give the same open models. The host that shows good test results can possibly get the customers.