Quali adds NVIDIA Nemotron 3 support to Torque platform

Quali has extended its Torque platform to support NVIDIA's Nemotron 3 family of AI models, broadening Torque's coverage across NVIDIA's NeMo and NIM software stack.

The platform now supports Nemotron 3 environments for inference, fine-tuning and reinforcement learning workloads across deskside systems, on-premises infrastructure and cloud GPU deployments.

This includes support for NVIDIA NIM microservices for inference and NeMo pipelines for training and model customisation. Torque is intended to provide policy enforcement, automated provisioning and lifecycle management for these environments as companies try to move AI projects beyond the pilot stage.

The announcement comes as NVIDIA rolls out the Nemotron 3 range in several tiers, including Nano, Super and Ultra, with associated tools for production inference and multi-agent system development. Torque is designed to manage those environments from initial deployment through shutdown.

Governance focus

Quali is positioning Torque as a control layer for businesses running AI workloads across a mix of hardware locations. The same blueprints and governance policies can be applied to systems such as NVIDIA DGX Spark and DGX Station, as well as larger on-premises GPU clusters and cloud infrastructure.

The platform integrates with NVIDIA GPU Operator and NIM Operator to provision and manage NIM microservice environments automatically. This allows teams to deploy inference services from versioned blueprints while applying internal rules and automated teardown processes.

For data scientists and machine learning engineers, Torque offers self-service access to Nemotron 3 fine-tuning and reinforcement learning environments through a portal interface. The setup includes cost attribution and compliance controls, aiming to reduce reliance on manual infrastructure requests.

Infrastructure control

For many organisations, the broader challenge is no longer access to AI models themselves, but managing the infrastructure needed to run them consistently and within internal policy limits. As AI systems become more complex, companies are placing greater focus on governance, resource use and operational oversight.

Quali's Nemotron 3 support builds on earlier work across NVIDIA's software and hardware ecosystem. Torque already supports DGX Spark, DGX Station, NVIDIA GPU Operator and NIM Operator, and now covers the full Nemotron 3 model family from Nano through Ultra.

Quali also said it is among the earliest software partners to validate Torque on NVIDIA's Blackwell architecture. That places the platform within a growing group of vendors trying to build management layers around increasingly specialised AI hardware and software stacks.

Executive view

Chief Executive Officer Lior Koriat outlined the company's view of the shift from experimentation to production use. "Nemotron 3 sets a new bar for open agentic AI, and enterprises are ready to put it to work. Torque gives them the governed infrastructure layer to do exactly that, deploying NIM services from versioned blueprints, running NeMo training environments that start and stop cleanly, and giving every team self-service access without infra complexity. This is how enterprises move from AI pilots to AI production," Koriat said.

Quali describes itself as an AI infrastructure management and Environment-as-a-Service automation provider. Its Torque platform is used to provision and govern AI workloads across GPU hardware, on-premises data centres, hybrid environments and public cloud systems from a single control plane.

The platform is aimed in part at regulated industries and government organisations that need technical enforcement of AI governance rules, including controls over who can access systems, how environments are configured and when resources are shut down.