tools
RunPod Pods
RunPod Pods are dedicated cloud GPU instances for AI development, training, inference, batch jobs, and long-running workloads.
RunPod Pods provide dedicated GPU environments with control over the GPU type, container, storage, drivers, and runtime. Users can deploy custom Docker images, run Linux GPU frameworks, and rent GPUs across 31 regions with billing by the second.
They are used by researchers, developers, and teams for model training, fine-tuning, inference, image and video generation, and AI agents. GPU compute is usage-priced, while container disks, volume disks, and network storage cost extra. Pods are separate from RunPod Serverless endpoints and multi-node Clusters.
Features
- Deploy dedicated GPU instances in under 30 seconds
- Choose from 30+ GPU models across 31 global regions
- Control containers, storage, GPU type, drivers, and runtime
- Run custom Docker images from registries
- Use per-second GPU billing with no long-term commitment
- Manage instances through an API, CLI, SDKs, and CI/CD
- Attach persistent or temporary storage
- Use spot instances at discounted rates with eviction risk
Use cases
- Run low-latency LLM inference with H100 or L40S instances
- Train and fine-tune models on A100 or H100 SXM GPUs
- Generate images and video with RTX 4090 or RTX A6000 GPUs
- Host stateful AI agents on persistent GPU instances
- Process batch jobs and other compute-heavy workloads
- Deploy custom containers with preferred ML frameworks
Pros
Cons
Pricing
- Starting price
- $0.27/hr
- Pricing checked
- 2026-09-19
B300
$7.89/hr
- 288 GB HBM3e
- 251 GB RAM
- 32 vCPUs
H200
$4.59/hr
- 141 GB VRAM
- 276 GB RAM
- 24 vCPUs
B200
$6.79/hr
- 180 GB VRAM
- 283 GB RAM
- 28 vCPUs
RTX Pro 6000
$2.09/hr
- 96 GB VRAM
- 188 GB RAM
- 16 vCPUs
H100 NVL
$3.19/hr
- 94 GB VRAM
- 94 GB RAM
- 16 vCPUs
H100 PCIe
$2.89/hr
- 80 GB VRAM
- 188 GB RAM
- 16 vCPUs