GPU pricing, without the mystery.
Two GPU fleets and serverless models, priced live. No hidden egress fees, no long-term contracts.
Two fleets with separate inventory and separate billing: QuickPods bills by the minute and can be fractionalised, TrainPods bills by the hour as whole GPUs. The Product column shows which fleet carries each card; the rate is the lowest across both.
| GPU | VRAM | Product | From / hr | Status | Action |
|---|---|---|---|---|---|
RTX 2000 Ada NVIDIA | 16 GB | TrainPods | ₹35 | Waitlist | |
RTX 4000 Ada NVIDIA | 20 GB | TrainPods | ₹41 | Waitlist | |
T4 NVIDIA 4 vCPU · 16 GB RAM | 16 GB | QuickPods | ₹49 | Available | Deploy |
L4 NVIDIA 8 vCPU · 24 GB RAM | 24 GB | Both | ₹56 | Available | Deploy |
A40 NVIDIA | 48 GB | TrainPods | ₹64 | Waitlist | |
RTX 3090 NVIDIA | 24 GB | TrainPods | ₹72 | Waitlist | |
RTX 4090 NVIDIA 8 vCPU · 32 GB RAM | 24 GB | QuickPods | ₹84 | Available | Deploy |
RTX 4090 NVIDIA | 24 GB | TrainPods | ₹99 | Available | Deploy |
V100 NVIDIA 8 vCPU · 32 GB RAM | 32 GB | QuickPods | ₹112 | Available | Deploy |
L40 NVIDIA | 48 GB | TrainPods | ₹118 | Waitlist | |
A100 NVIDIA | 40 GB | TrainPods | ₹128 | Available | Deploy |
L40S NVIDIA | 48 GB | Both | ₹140 | Available · 2 | Deploy |
A30 NVIDIA | 24 GB | TrainPods | ₹193 | Waitlist | |
A100 NVIDIA 16 vCPU · 64 GB RAM | 80 GB | Both | ₹198 | Available | Deploy |
H100 PCIe NVIDIA | 80 GB | TrainPods | ₹239 | Waitlist | |
H100 NVIDIA 32 vCPU · 128 GB RAM | 80 GB | QuickPods | ₹349 | Available | Deploy |
H100 SXM NVIDIA | 80 GB | TrainPods | ₹384 | Waitlist | |
H100 NVL NVIDIA | 94 GB | TrainPods | ₹455 | Waitlist | |
H200 NVL NVIDIA | 141 GB | TrainPods | ₹493 | Waitlist | |
H200 SXM NVIDIA | 141 GB | TrainPods | ₹569 | Waitlist | |
B200 NVIDIA | 192 GB | TrainPods | ₹840 | Waitlist | |
B300 NVIDIA | 288 GB | TrainPods | ₹1,053 | Waitlist |
RTX 2000 Ada
NVIDIA · 16 GB VRAM
RTX 4000 Ada
NVIDIA · 20 GB VRAM
A40
NVIDIA · 48 GB VRAM
RTX 3090
NVIDIA · 24 GB VRAM
L40
NVIDIA · 48 GB VRAM
A30
NVIDIA · 24 GB VRAM
H100 PCIe
NVIDIA · 80 GB VRAM
H100 SXM
NVIDIA · 80 GB VRAM
H100 NVL
NVIDIA · 94 GB VRAM
H200 NVL
NVIDIA · 141 GB VRAM
H200 SXM
NVIDIA · 141 GB VRAM
B200
NVIDIA · 192 GB VRAM
B300
NVIDIA · 288 GB VRAM
Rates shown are for a whole GPU. QuickPods can be fractionalised — you can take a slice of a card rather than the whole thing, and pay the matching fraction of the hourly rate. TrainPods offerings are whole GPUs.
Starting prices are the lowest available configuration. Instances of the same GPU vary in price with the vCPU, RAM and storage attached to them, so the rate you see at launch depends on the configuration you pick.
One-click templates billed by the minute, on a fleet you can take a fraction of. You pay for GPU time only — the templates are free.
21 templates across 7 categories
LLM & Fine-tuning
4- Axolotl CUDA 12
- LLaMA Factory CUDA 12
- Unsloth CUDA 13
- Unsloth Studio Cuda 13
LLM Inference
6- NVIDIA Triton CUDA 12
- Ollama CUDA 12
- SD WebUI Forge CUDA 12
- TensorRT CUDA 12
- +2 more
Image Generation
4- Automatic1111 CUDA 12
- ComfyUI Cuda12
- Kohya-ss CUDA 12
- Meta SAM3
ML Frameworks
2- NVIDIA RAPIDS CUDA 12
- PyTorch Cuda12 OpenCV
Scientific Computing
3- AlphaFold CUDA 12
- GNU Octave CUDA 12
- GROMACS CUDA 12
3D & Rendering
1- FFmpeg GPU CUDA 12
Development Tools
1- JupyterLab GPU CUDA 12
Serverless model endpoints, billed per token rather than per GPU-hour. No idle cost between requests.
12 language models · billed per 1M tokens
| Model | Context | Input / 1M | Output / 1M |
|---|---|---|---|
| NVIDIA: Nemotron 3 Ultra (free) | 1M | Free | |
| Microsoft Phi 4 | 16K | ₹9.42 | ₹19 |
| NVIDIA Nemotron 3 Super | 1M | ₹11 | ₹54 |
| Llama 4 Scout | 10M | ₹13 | ₹40 |
| Qwen: Qwen3 Coder Next | 262K | ₹15 | ₹108 |
| DeepSeek V4 Flash | 1M | ₹19 | ₹38 |
| Llama 4 Maverick | 1M | ₹27 | ₹108 |
| Mistral Devstral 2 2512 | 262K | ₹54 | ₹269 |
| Xiaomi MiMo V2.5 Pro | 1M | ₹59 | ₹117 |
| MoonshotAI Kimi K2 0711 | 131K | ₹77 | ₹309 |
| GLM 5.2 | 1M | ₹86 | ₹271 |
| MoonshotAI Kimi Latest | 262K | ₹404 | ₹2,018 |
Other modalities · billed differently
Sam 3
videoLTX Video 2.3 (Text-to-Video)
videoThe smallest billable unit is a minute, not an hour. TrainPods bills by the hour.
Move your data out without surprise charges.
Right-size from 12.5% of a card up to 8x full GPUs. TrainPods offerings are whole GPUs.