Runpod is the GPU cloud I render my AI videos on, including my 100 day videos and Nero clips.
Runpod's pricing page is long. It lists more than 20 GPUs, two clouds, Serverless workers, Clusters, storage and ready made AI model endpoints, each priced a different way. This post puts it in plain English so you can work out what a job will really cost.
Every number below comes from Runpod's official pricing page, which was marked "Updated November 9, 2026" when I checked it on November 11, 2026. Prices change often, so check the page before you start a big job.
The quick answer
- Pods are GPUs you rent by the hour, billed by the second with no commitment. Prices on the page ran from $0.16 an hour (RTX A5000 on Community Cloud) to $8.99 an hour (B300 on Secure Cloud).
- Community Cloud is cheaper than Secure Cloud for the same GPU. For example, an RTX 4090 was $0.34 an hour on Community Cloud and $0.89 on Secure Cloud.
- Serverless bills workers only while they handle requests, at a higher hourly rate than Pods.
- Clusters rent several GPUs together, priced per GPU per hour, up to 64 GPUs.
- Storage is extra, from $0.05 per GB a month, and an idle volume disk costs more than a running one.
How Runpod pricing works
Runpod sells compute in a few ways:
1. Pods: a dedicated GPU instance (a container) that you start, use and stop. Best for rendering, training, notebooks and long jobs. 2. Serverless: you deploy a model as an endpoint, and workers spin up to handle requests. Best for inference and APIs with uneven traffic. 3. Clusters: multi GPU, multi node jobs you can launch in minutes, with no commitment. 4. Reserved Clusters: dedicated capacity on 1 to 12 month or longer terms, priced by Runpod's sales team. 5. Public Endpoints: ready made AI models (image, video, audio and language) you pay for per request, per character or per token.
Pod GPU prices (per hour)
Here's what the pricing page listed for on-demand Pods. Secure Cloud runs in Runpod's partner data centres; Community Cloud runs on vetted third party hosts.
| GPU | VRAM | Secure Cloud | Community Cloud |
|---|---|---|---|
| B300 | 288 GB | $8.99 | $6.94 |
| B200 | 180 GB | $7.99 | $5.98 |
| H200 | 141 GB | $5.29 | $3.59 |
| H100 SXM | 80 GB | $3.99 | $2.69 |
| H100 NVL | 94 GB | $3.19 | $2.59 |
| H100 PCIe | 80 GB | $2.89 | $1.99 |
| RTX Pro 6000 | 96 GB | $2.49 | $1.69 |
| A100 SXM | 80 GB | $1.79 | $1.39 |
| A100 PCIe | 80 GB | $1.79 | $1.19 |
| Pro 6000 MIG 48GB | 48 GB | $1.29 | $1.00 |
| RTX 5090 | 32 GB | $1.19 | $0.69 |
| L40S | 48 GB | $1.09 | $0.79 |
| RTX 6000 Ada | 48 GB | $0.99 | $0.74 |
| RTX 4090 | 24 GB | $0.89 | $0.34 |
| L40 | 48 GB | $0.82 | $0.69 |
| Pro 6000 MIG 24GB | 24 GB | $0.69 | $0.50 |
| A40 | 48 GB | $0.59 | $0.35 |
| RTX A6000 | 48 GB | $0.59 | $0.33 |
| L4 | 24 GB | $0.59 | $0.44 |
| RTX 3090 | 24 GB | $0.50 | $0.22 |
| RTX A5000 | 24 GB | $0.27 | $0.16 |
Prices are in US dollars per GPU hour. Runpod's FAQ says Pods are "billed per second with no commitment", so a 20 minute job is charged for 20 minutes, not a full hour.
What the table means in practice: for most image and video generation, a 24 GB card (RTX 4090 or RTX 3090) or a 32 GB RTX 5090 is the usual starting point. You only need the 80 GB and larger cards for big models or training. The gap between the clouds is biggest on consumer cards: the RTX 4090 was less than half the price on Community Cloud.
Secure Cloud or Community Cloud?
- Secure Cloud costs more, and runs in Runpod's partner data centres. Pick it for client work, sensitive data or anything that must stay up.
- Community Cloud is cheaper, and runs on vetted third party hosts. It's a good fit for testing, personal projects and jobs you can restart.
A simple rule: try your workflow on Community Cloud, then move to Secure Cloud when it matters more that it runs without interruption than that it's cheap.
Serverless prices (per hour of worker time)
Serverless is priced by GPU size. You pay for the time your workers are running, not for a Pod sitting idle. Here's what the page listed:
| GPU size | GPUs | Price per hour |
|---|---|---|
| 288 GB | B300 | $11.99 |
| 180 GB | B200 | $10.79 |
| 141 GB | H200 | $6.99 |
| 96 GB | RTX 6000 Pro | $3.49 |
| 80 GB | H100 | $5.39 |
| 80 GB | A100 | $2.72 |
| 48 GB | L40, L40S, 6000 Ada, MIG 48GB | $1.75 |
| 48 GB | A6000, A40 | $1.22 |
| 32 GB | RTX 5090 | $1.79 |
| 32 GB | RTX PRO 4500 Blackwell | $1.15 |
| 24 GB | RTX 4090 | $1.34 |
| 24 GB | L4, A5000, 3090, MIG 24GB | $0.75 |
| 16 GB | A4000, A4500, RTX 4000, RTX 2000 | $0.58 |
Serverless costs more per hour than the same GPU as a Pod, so it only saves money when your traffic comes and goes. If a model is busy all day, a Pod is usually cheaper. Runpod also says Serverless reservations and contract pricing are available through its sales team.
Cluster prices (per GPU per hour)
| GPU | Price per GPU hour |
|---|---|
| B300 | $10.79 |
| B200 | $9.59 |
| H200 SXM | $6.35 |
| H100 SXM | $4.79 |
| A100 SXM | $2.15 |
Instant Clusters scale up to 64 GPUs with shared storage and no commitment. Reserved Clusters (B200, H200 SXM, H100 SXM, A100 SXM, B300 and L40S) are priced by contacting sales.
Storage costs
This is the part people forget. Storage is billed per GB per month on top of GPU time:
| Storage type | Price |
|---|---|
| Container disk | $0.10 per GB a month |
| Volume disk, while the Pod is running | $0.10 per GB a month |
| Volume disk, while the Pod is stopped (idle) | $0.20 per GB a month |
| Network storage (standard), under 1 TB | $0.07 per GB a month |
| Network storage (standard), over 1 TB | $0.05 per GB a month |
| Network storage (high performance) | $0.14 per GB a month |
A quick example: a stopped Pod with a 100 GB volume disk costs about $20 a month at $0.20 per GB, even though no GPU is running. Delete volumes you don't need, or move files you want to keep to network storage.
Public Endpoints
If you just want to call a model without setting anything up, Runpod lists ready made endpoints with their own prices. A few examples from the page: FLUX.1 Schnell images at $0.0024 per megapixel, FLUX.1 [dev] at $0.02 per megapixel, Wan 2.2 image to video at $0.30 per 5 second 720p clip, and Minimax Speech 02 HD at $0.05 per 1,000 characters. The full list is long and changes often, so check the page for the model you want.
How to keep your Runpod bill down
- Stop Pods when you're done. Billing is per second, so stopping straight away saves money. Remember stopped volume disks still cost $0.20 per GB a month.
- Start on Community Cloud for tests, then move to Secure Cloud for important work.
- Pick the smallest GPU that fits your model. A 24 GB card at under $1 an hour is enough for a lot of image and video work.
- Use Serverless only for uneven traffic. For steady, all day use, a Pod is cheaper per hour.
- Watch your storage. Clear out old models and outputs you no longer need.
Try Runpod
You only pay for what you use, with no commitment.
๐ Try Runpod here (Paid Link)
Reminder: that's my affiliate link. If you sign up through it and spend on Runpod, I may earn a commission, at no extra cost to you.
FAQ
How much does Runpod cost per hour?
It depends on the GPU and cloud. On the pricing page I checked, Pods ran from $0.16 an hour (RTX A5000, Community Cloud) to $8.99 an hour (B300, Secure Cloud). An RTX 4090 was $0.34 an hour on Community Cloud and $0.89 on Secure Cloud.
Does Runpod charge by the second?
Yes. Runpod's FAQ says Pods are billed per second with no commitment.
How much is an H100 on Runpod?
The H100 PCIe was $2.89 an hour on Secure Cloud or from $1.99 on Community Cloud. The H100 SXM was $3.99 on Secure Cloud or from $2.69 on Community Cloud.
Do I pay for storage when my Pod is stopped?
Yes. A volume disk costs $0.20 per GB a month while the Pod is idle, compared with $0.10 while it's running.
Is Runpod Serverless cheaper than Pods?
Not per hour. Serverless rates are higher than the same GPU as a Pod, but you only pay while workers run, so it can be cheaper when traffic is uneven.
Does Runpod have a free tier?
The pricing page doesn't list a free tier. You pay as you go for what you use.
Sources
Checked on November 11, 2026 (page marked "Updated November 9, 2026"). Prices change, so check the official page before you buy.
- Runpod pricing: https://www.runpod.io/pricing