NVIDIA Run:ai

Open source

GPU orchestration platform that pools and schedules AI workloads across Kubernetes-based clusters.

Pricing: free, custom
License: open-source
Open source: Yes

Editorial review

NVIDIA Run:ai has not yet been editorially reviewed. User reviews and published field data below are unaffected — we say so rather than implying an editorial verdict.

Facts & attributes

Every value carries its evidence state. “—” means unknown — never a silent default (P-02).

Identity · Taxonomy

Documentation(HTTPS URL): https://docs.nvidia.com/run-ai/index.htmlVerified
Tags(open-source;cloud-native): GPU orchestration, Kubernetes, AI workloads, open source, MLOpsVerified

Platform · Deployment

Delivery model(SaaS;self-hosted;desktop;mobile;embedded;managed-cloud;appliance): self-hostedVerification pending

Pricing · Licensing

License model(proprietary;open-source;source-available;dual-license): open-sourceVerification pending
Pricing model(free;subscription;usage;seat;asset;transaction;revenue-share;perpetual;custom): free, customVerification pending
Source repository(HTTPS URL): https://github.com/dsx-ai-factory/run-aiVerified

Features · Product · Fit

Core capabilities(authoring;automation;analytics): AI-native workload orchestration for training, fine-tuning, and inference, Dynamic GPU allocation to reduce idle or fragmented capacity, Policy-driven governance with quotas, priorities, and access controls, Fractional GPU allocations with KAI GPU Fractioning Operator, KAI Scheduler with hierarchical queues, fair-share, and topology-aware placement, Model Streamer for streaming tensors toward GPU memoryVerification pending

Api · Integrations · Ecosystem

API available(true/false): trueVerification pending

Audience · Use · Case

Best for(Neutral editorial summary): Enterprises and platform teams orchestrating AI training and inference on shared GPU infrastructure at scaleVerification pending
Company-size fit(individual;SMB;mid-market;enterprise): enterpriseVerified
Primary use cases(build;monitor;automate): AI training, fine-tuning, and inference at scale, Governing shared GPU infrastructure, Enterprise AI acceleration, Maximizing token throughput, Mitigating model cold startVerification pending

Vendor · Maintainer

Founded year(2012): 1993Verified
Headquarters country(US): United StatesVerified
Vendor type(commercial;foundation;community;government;consortium;individual): commercialVerified

User reviews

Review policy

No published reviews yet.