Model APIs
Token Factory
Call hosted open models through an OpenAI-compatible API. Pay per million tokens.
- Input
- $0.13
- Cached
- $0.03
- Output
- $0.26
- Input
- $0.07
- Cached
- $0.003
- Output
- $0.49
- Input
- $0.11
- Cached
- $0.05
- Output
- $0.37
- Input
- $0.14
- Cached
- $0.05
- Output
- $0.40
- Input
- $1.21
- Cached
- $0.22
- Output
- $3.86
- Input
- $0.14
- Cached
- $0.03
- Output
- $0.45
- Input
- $0.15
- Cached
- $0.02
- Output
- $0.50
- Input
- $0.30
- Cached
- $0.08
- Output
- $2.35
Rates in USD per 1 million tokens. Input is the context you send; output is what the model generates. Cached input rates apply to verified cache hits. Sign in to see prices in your billing currency. Learn more about Token Factory →
Compute
GPUs
NVIDIA H200, H100, RTX Pro 6000 Blackwell, A100, A30 and L4. Pick 1, 2, 4 or 8 GPUs to price an instance.
NVIDIA H200 SXM
NVIDIA H100 SXM
NVIDIA RTX Pro 6000 Blackwell
NVIDIA A100 80GB
NVIDIA A100 40GB
NVIDIA A30
NVIDIA L4
NVIDIA H200 SXM
141 GB VRAM · 28 vCPU · 300 GB RAM
- On-demand / hr
- $4.59
- Spot / hr
- $2.29−50%
NVIDIA H100 SXM
80 GB VRAM · 16 vCPU · 200 GB RAM
- On-demand / hr
- $3.49
- Spot / hr
- $1.54−56%
NVIDIA RTX Pro 6000 Blackwell
96 GB VRAM · 28 vCPU · 160 GB RAM
- On-demand / hr
- $2.19
- Spot / hr
- $1.15−47%
NVIDIA A100 80GB
80 GB VRAM · 16 vCPU · 112 GB RAM
- On-demand / hr
- $1.49
- Spot / hr
- $0.89−40%
NVIDIA A100 40GB
40 GB VRAM · 16 vCPU · 112 GB RAM
- On-demand / hr
- $0.89
- Spot / hr
- $0.79−11%
NVIDIA A30
24 GB VRAM · 16 vCPU · 112 GB RAM
- On-demand / hr
- $0.41
- Spot / hr
- $0.29−29%
NVIDIA L4
24 GB VRAM · 32 vCPU · 124 GB RAM
- On-demand / hr
- $0.44
- Spot / hr
- $0.29−34%
On-demand instances are billed per minute; pause and resume any time. Spot instances cost less but can be interrupted when the capacity is needed, so save your work often. Customers outside India are billed in USD. Switch at the top right to see our India rates. Committing to a month or more? Talk to sales about a savings plan →
CPU VMs
General-purpose VMs with 4 GB of RAM per vCPU. On-demand only.
2 vCPU · 8 GB
4 vCPU · 16 GB
8 vCPU · 32 GB
16 vCPU · 64 GB
32 vCPU · 128 GB
2 vCPU · 8 GB RAM
4 vCPU · 16 GB RAM
8 vCPU · 32 GB RAM
16 vCPU · 64 GB RAM
32 vCPU · 128 GB RAM
CPU VMs are priced as $0.012 / vCPU + $0.0032 / GB RAM per hour. Available in the India (Noida) region.
GPU Clusters
Multi-node training over InfiniBand.
Instant Clusters
NVIDIA H200 SXM nodes, eight GPUs each, over InfiniBand. Create one yourself from the dashboard.
Reserved Clusters
128 to 1,024+ GPUs for a planned project. We confirm the configuration, dates and pricing with you.
Storage
One rate for instance disks and filesystems.
Disks & filesystems
Instance disks and network filesystems, billed by the hour on provisioned capacity.
Egress
Moving data out of your instances within the region.
Deploy
Deploy your own model
Run a model on dedicated GPU workers. Billed by the minute of worker runtime.
Managed Endpoints
Deploy a catalog model on a serving stack we maintain. Warm workers are billed even when idle.
Serverless
Deploy your own model and serving stack. Workers scale with traffic; set minimum workers to zero to stop when idle.
The GPU rate for a deployment is shown in the dashboard before you deploy. Each worker also has a 50 GB root disk, and any filesystem holding model weights is billed at the storage rate above.
Need 25+ GPUs or a savings plan?
Discounted rates for one-month and longer commitments, reserved capacity, NVIDIA H200 InfiniBand clusters and custom quotes. Talk to us about a plan that fits your workload.
Trusted by 27,000+ AI developers · 50M+ GPU hours served
What builders say
“I published a paper with the results from exps i did on your gpus when @googlecloud turned me down. Thanks a lot for the great product/service. Not to mention super competitive prices...”
“After waiting for 3 days on AWS to provide capacity so I can continue on @fastdotai course, I switched to @jarvislabsai and it literally took me less than 5 min. Good stuff @vishnuvig”
“JarvisLabs.ai has always been my go to for renting GPU for personal projects/Kaggle competitions. Trust me, it made my life easier. Kudos Vishnu for your dedication, and good luck for the future!”
“I have had a great experience using the Jarvislabs platform. Within a couple of minutes you can have a machine up with multiple connection/interaction options like Jupyter, VS Code, and normal SSH”
