Best GPU AI tools
Accelerated computing on GPUs, from CUDA workloads to rented inference and training capacity. The 11 highest-ranked of the 11 apps tagged GPU in the Onei AI catalogue.
Everything tagged GPU: apps, skills and MCP servers Best GPU agent skills, ranked
-
Unsloth AI
Open-source desktop app to run, serve and fine-tune text, image, video and audio models entirely on your own machine.
Also: Fine-tuning, Inference · Editor's pick
-
A generative media platform tuned for speed — image, video, and audio models served with very low latency.
Also: API, Developers, Inference
-
Run thousands of open models with one API call, and deploy your own without touching Kubernetes.
Also: API, Developers, Hosting
-
Inference and training platform for open models: serverless per-token APIs, dedicated GPU deployments, and managed fine-tuning with SFT, DPO and RFT.
Also: API, Developers, Fine-tuning
-
GPU cloud for AI developers: on-demand GPU pods, serverless GPU endpoints that scale to zero, and multi-node clusters, billed by the second.
Also: Developers, Fine-tuning, Inference
-
NVIDIA GPU cloud for training and inference: self-serve on-demand instances billed by the minute, 1-Click Clusters, and large private superclusters.
Also: Developers, Inference
-
Simulate how fast a local LLM runs on a given GPU or Mac before you buy the hardware.
Also: Inference
-
Production inference for open and custom models — fast, autoscaling, and deployable into your own cloud.
Also: Developers, Hosting, Inference
-
Serverless GPUs from a Python decorator — deploy models and batch jobs with no containers or cluster to manage.
Also: Developers, Hosting, Inference
-
Inference, fine-tuning, and GPU clusters for open models — the full stack for teams building on open weights.
Also: API, Developers, Fine-tuning
-
Inference on custom LPU hardware, built for latency — open models served at speeds general-purpose GPUs struggle to match.
Also: API, Developers, Inference
How this list is ranked
Ranked by editorial weight (0–100), plus 5 points per bookmark up to 10, plus or minus up to 20 for the average review rating; ties go to featured apps, then the most recently updated. Discontinued apps are never ranked.