Inference platformsOnline
Together AI
AI platform offering serverless and dedicated inference APIs for open models, fine-tuning, and self-serve GPU clusters (H100, H200, B200, B300, GB200/GB300)…
- Lowest on-demand H100 rate
- 399 US cents/GPU-h
Compare inference platforms
Check service regions with each provider.
Results
31 GPU cloud providers
Inference platformsOnline
Atlas Cloud
Atlas Cloud offers a single OpenAI-compatible API for 400+ image, video, audio and language models, plus on-demand GPU instances with NVIDIA H100 SXM and H200…
- Lowest on-demand H100 rate
- 295 US cents/GPU-h
Inference platformsOnline
Baseten
AI inference platform offering model APIs, dedicated model deployments and training on GPU instances from T4 to B200, billed by the minute, with options to…
- Lowest on-demand H100 rate
- 650 US cents/GPU-h
- Lowest on-demand A100 rate
- 400 US cents/GPU-h
Serverless GPUOnline
Cerebrium
Serverless AI infrastructure platform for deploying LLM, voice and image workloads on GPUs from T4 to B200, scaling with demand and charging for compute time…
- Lowest on-demand H100 rate
- 340 US cents/GPU-h
- Lowest on-demand A100 rate
- 200 US cents/GPU-h
GPU neocloudsOnline
Crusoe
AI cloud offering NVIDIA and AMD GPU instances (GB300, B300, H200, H100, A100, L40S, MI300X) on spot, on-demand or reserved terms, plus managed Kubernetes,…
- Lowest on-demand H100 rate
- 390 US cents/GPU-h
- Lowest on-demand A100 rate
- 200 US cents/GPU-h
Inference platformsOnline
DeepInfra
Inference platform serving open and proprietary models via API priced per token or per unit, plus custom model deployments on dedicated A100, H100, H200, B200…
- Lowest on-demand H100 rate
- 290 US cents/GPU-h
- Lowest on-demand A100 rate
- 149 US cents/GPU-h
Inference platformsOnline
Featherless
Serverless inference platform for open-weight models with an OpenAI-compatible API, monthly chat subscriptions, metered developer plans and dedicated GPU…
Inference platformsOnline
Fireworks AI
Generative AI inference platform offering serverless per-token model APIs, fine-tuning and training, and on-demand GPU deployments on H100 to GB300 GPUs priced…
- Lowest on-demand H100 rate
- 800 US cents/GPU-h
Inference platformsOnline
FriendliAI
AI inference platform offering model APIs, on-demand dedicated GPU endpoints on A100 to B300 GPUs billed per second, and bring-your-own-GPU deployments, with…
- Lowest on-demand H100 rate
- 500 US cents/GPU-h
- Lowest on-demand A100 rate
- 400 US cents/GPU-h
GPU neocloudsOnline
GMI Cloud
GMI Cloud provides dedicated single-tenant NVIDIA GPUs (H100, H200, B200, GB200, GB300) in GMI-operated data centers, on-demand or reserved, priced per…
- Lowest on-demand H100 rate
- 260 US cents/GPU-h
Inference platformsOnline
Hugging Face Inference Endpoints
Managed service from Hugging Face for deploying models on dedicated, autoscaling endpoints on AWS, GCP or Azure infrastructure, with GPU instances (T4, L4,…
- Lowest on-demand H100 rate
- 1,000 US cents/GPU-h
- Lowest on-demand A100 rate
- 250 US cents/GPU-h
Serverless GPUOnline
Modal
Serverless AI infrastructure platform for running inference, training, batch jobs and sandboxes on GPUs and CPUs that spin up in seconds, billed per second for…
- Lowest on-demand H100 rate
- 395 US cents/GPU-h
- Lowest on-demand A100 rate
- 210 US cents/GPU-h
Inference platformsOnline
Novita AI
AI cloud offering model inference APIs for LLM, image, video and audio, plus GPU instances and serverless GPUs across 14 global regions, with on-demand and…
Inference platformsOnline
Parasail
AI inference platform offering per-token serverless endpoints for open models, self-service batch processing, and dedicated deployments on reserved NVIDIA GPUs…
GPU neocloudsOnline
Phala Cloud (GPU TEE)
Confidential AI cloud running containers, private LLM inference and GPU jobs inside hardware TEEs (Intel TDX, NVIDIA confidential computing) with attestation;…
- Lowest on-demand H100 rate
- 308 US cents/GPU-h
Inference platformsOnline
Segmind
Generative AI platform offering serverless pay-as-you-go APIs for image, video and other media models, Pixelflow workflows, fine-tuning, and dedicated GPU…
- Lowest on-demand H100 rate
- 900 US cents/GPU-h
- Lowest on-demand A100 rate
- 432 US cents/GPU-h
Inference platformsOnline
Anyscale
Managed Ray platform for distributed training, batch inference and model serving, offering Anyscale-hosted GPU compute (T4, L4, A10G, A100, H/B/GB families) or…
Inference platformsOnline
fal
Generative media platform offering model APIs for image, video, audio and 3D generation billed per output, plus serverless compute for deploying custom apps on…
- Lowest on-demand H100 rate
- 450 US cents/GPU-h
GPU marketplacesOnline
Hyperbolic
AI compute platform offering on-demand and reserved NVIDIA H100, H200 and B200 GPUs sourced from a supplier network, managed Kubernetes with GPU worker nodes,…
- Lowest on-demand H100 rate
- 319 US cents/GPU-h
Inference platformsOnline
Inference.net
Inference infrastructure for open-source, custom and fine-tuned models, offering dedicated model deployments on B200 GPU instances priced per hour, plus…
Decentralized GPUOnline
io.net
Decentralized GPU cloud aggregating GPUs from many suppliers; offers on-demand GPU clusters (Ray, Kubernetes, containers, bare metal) plus io.intelligence…
- Lowest on-demand H100 rate
- 219 US cents/GPU-h
GPU marketplacesOnline
Prime Intellect
AI training, inference and compute platform offering on-demand access to 1-256 GPUs across clouds, reserved clusters quoted from 50+ providers, a spot resale…
Inference platformsOnline
Super Protocol
Confidential AI platform: Super Swarm runs AI workloads in CPU/GPU TEEs across clouds and data centers, and Confidential Router serves open models via an…
Decentralized GPUOnline
Theta EdgeCloud
Hybrid cloud-edge platform from Theta offering on-demand GPU compute, hosted AI model inference, AI agents and video/rendering developer tools; contributors…
- Lowest on-demand H100 rate
- 229 US cents/GPU-h
- Lowest on-demand A100 rate
- 199 US cents/GPU-h
Other categories in Online
More GPU cloud providers
Questions
What other kinds of GPU cloud providers are in Online?
Other categories in Online: GPU neoclouds (34 GPU cloud providers), Bare-metal GPU clusters (20 GPU cloud providers), GPU marketplaces (15 GPU cloud providers), Serverless GPU (15 GPU cloud providers), and Decentralized GPU (9 GPU cloud providers).
Published by GPUClouds. How we check each GPU cloud provider.