Pricing
GPU time and API requests
Explore for free. When you’re ready to run, choose the setup that fits your work and review the rate before launching.
Credits pay for usage. Creating an account does not launch a machine.
Dedicated GPU Instances
Choose compute reserved for your workload. Running time, including startup and idle time, is billed. Stopped machines may retain billable storage.
| GPU | VRAM | vCPU | RAM | Price/hr | Typical uses |
|---|---|---|---|---|---|
| CPU Only | — | 3 | 16 GB | $0.15/hour | Small LLMs (1B-3B), basic inference, API servers, testing |
| RTX 4090 | 24 GB | 6 | 41 GB | $0.99/hour | 7B-13B LLMs, SDXL Images, Flux, Most Popular GPU |
| L4 | 24 GB | 12 | 50 GB | $0.66/hour | 7B-13B LLMs, SDXL Images, Power Efficient |
| RTX A6000 | 48 GB | 9 | 50 GB | $0.72/hour | 30B+ LLMs (quantized 70B), Video Gen, 48GB VRAM |
| A100 80GB | 80 GB | 8 | 117 GB | $1.85/hour | 70B+ LLMs, Complex Workflows, High-VRAM Inference |
| H100 | 80 GB | 16 | 188 GB | $4.32/hour | Lowest-Latency Inference, Large Models, Complex Workflows |
| RTX 6000 Ada Generation | 48 GB | 12 | 62 GB | $1.12/hour | Video Gen, MiniMax H3, FP8/INT8 quantized models, 48GB VRAM |
| L40S | 48 GB | 12 | 62 GB | $1.31/hour | Video Gen, MiniMax H3, high-bandwidth inference, 48GB VRAM |
| RTX PRO 6000 Blackwell Server Edition | 96 GB | 16 | 188 GB | $2.76/hour | MiniMax H3, NVFP4/FP8 quantized models, 2K video, 96GB VRAM |
CPU Only
$0.15/hourSmall LLMs (1B-3B), basic inference, API servers, testing
RTX 4090
$0.99/hour7B-13B LLMs, SDXL Images, Flux, Most Popular GPU
L4
$0.66/hour7B-13B LLMs, SDXL Images, Power Efficient
RTX A6000
$0.72/hour30B+ LLMs (quantized 70B), Video Gen, 48GB VRAM
A100 80GB
$1.85/hour70B+ LLMs, Complex Workflows, High-VRAM Inference
H100
$4.32/hourLowest-Latency Inference, Large Models, Complex Workflows
RTX 6000 Ada Generation
$1.12/hourVideo Gen, MiniMax H3, FP8/INT8 quantized models, 48GB VRAM
L40S
$1.31/hourVideo Gen, MiniMax H3, high-bandwidth inference, 48GB VRAM
RTX PRO 6000 Blackwell Server Edition
$2.76/hourMiniMax H3, NVFP4/FP8 quantized models, 2K video, 96GB VRAM
All prices include compute, memory, disk, and network.
Serverless — Pay Per Request
Review supported models and per-request prices. First requests may wait for startup; text chat needs a configured endpoint.
Text Models
| Model | Per Request |
|---|---|
| Qwen3 4BQuick Response | $0.014 |
| Qwen3 8BGeneral Purpose | $0.019 |
| DeepSeek R1 8BReasoning | $0.019 |
| GLM-Z1 9BBilingual Reasoning | $0.019 |
| DeepSeek R1 14BAdvanced Reasoning | $0.047 |
| Qwen3 32BPremium Quality | $0.138 |
| GLM-Z1 32BPremium Reasoning | $0.138 |
Image Models
| Model | Per Request |
|---|---|
| Flux Schnell | $0.017 |
| Flux Dev | $0.017 |
| Stable Diffusion XL | $0.014 |
| Z-Image Turbo | $0.014 |
Video runs use dedicated workflows. Explore video workflows
Generation APIs use prepaid credits per successful request. Text chat requires a running deployment or configured endpoint; availability is checked when a request runs.
Estimate Your Costs
Daily
$2.64
Monthly
$57
What can I do before paying?
Browse the model catalog, inspect available workflows, and use the workflow checker without an account. When available, the free demo lets you try a supported model within its usage limits.
How do dedicated machines differ from serverless?
A dedicated machine is reserved for your workload and is billed by time. Serverless runs supported models per request. Their available models, configuration options and prices differ; compare the tables above.
What happens when I finish using a machine?
Manage dedicated machines from your dashboard. Closing the browser does not remove a machine. Stopping and removing are different actions: stopped machines can retain storage, so review the resource details and remove resources you no longer need.
Is the calculator my final bill?
No. It estimates use at the listed rate. Setup, storage, availability and actual usage depend on your configuration. Review the deployment details before launching and your recorded usage in Billing afterward.
Do I need a subscription?
Dedicated deployments use prepaid credits. You can review the available credit amounts in your account before checking out.
Workflow guides
Start with a workflow.
See its requirements before choosing a machine.