GPU Specs API

Free REST API with specs for over 12,013 NVIDIA, AMD and Intel GPUs. JSON format, no registration, no API key required.

Base URL

https://gpuark.com/api/v1/

Endpoints

GET /api/v1/gpus/

List of all GPUs with pagination.

Query parameters:

Parameter	Type	Description
page	int	Page number (default 1)
per_page	int	Items per page (default 50, max 200)
vendor	string	Filter by manufacturer: `nvidia`, `amd`, `intel`
search	string	Search by GPU name
sort	string	Sort by: `gpi`, `name`, `ram`, `cores`, `release_date`. Prefix `-` for descending
min_gpi	float	Minimum GPU Performance Index value

Example request:

curl "https://gpuark.com/api/v1/gpus/?vendor=nvidia&sort=-gpi&per_page=5"

Example response:

{
  "count": 1842,
  "page": 1,
  "per_page": 5,
  "total_pages": 369,
  "results": [
    {
      "name": "NVIDIA H100 SXM5 80 GB",
      "slug": "nvidia-h100-sxm5-80-gb",
      "vendor": "NVIDIA",
      "manufacturer": "NVIDIA",
      "ram": 80,
      "ram_type": "HBM3",
      "cores": 16896,
      "gpi_value": 187.5,
      "fp32_performance": 66.91,
      "card_release_date": "2023-03-21",
      "url": "https://gpuark.com/ru/gpu/nvidia-h100-sxm5-80-gb/"
    }
  ]
}

GET /api/v1/gpus/{slug}/

Full specifications for a specific GPU.

Example:

curl "https://gpuark.com/api/v1/gpus/nvidia-geforce-rtx-4090/"

Returns all specifications: VRAM, clock speeds, performance (FP16/FP32/FP64/BF16/TF32/INT8), bus, TDP, API (CUDA, Vulkan, OpenCL), NVENC/NVDEC and GPI.

GET /api/v1/compare/{slug1}-vs-{slug2}/

Comparison of two GPUs.

Example:

curl "https://gpuark.com/api/v1/compare/nvidia-geforce-rtx-4090-vs-nvidia-geforce-rtx-3090/"

Rate Limits

No registration or API key required
Up to 100 requests per minute per IP
Data is cached for 1 hour
Maximum 200 items per page

Usage Examples

Python:

import requests

# All NVIDIA GPUs with GPI > 100
response = requests.get("https://gpuark.com/api/v1/gpus/", params={
    "vendor": "nvidia",
    "min_gpi": 100,
    "sort": "-gpi",
})
gpus = response.json()["results"]

for gpu in gpus:
    print(f"{gpu['name']}: GPI {gpu['gpi_value']}, {gpu['ram']} GB {gpu['ram_type']}")

JavaScript:

// Specifications for a Specific GPU
const res = await fetch("https://gpuark.com/api/v1/gpus/nvidia-geforce-rtx-4090/");
const gpu = await res.json();
console.log(`${gpu.name}: ${gpu.fp32_performance} TFLOPS, ${gpu.ram} GB ${gpu.ram_type}`);

Embeddable Widgets

Ready-made widgets for embedding on your site via <iframe>. Data is always up to date, GPU Ark attribution is included automatically.

GPU Card

Compact block with key specifications for a single GPU. Suitable for reviews and articles.

Embed code:

<iframe
  src="https://gpuark.com/en/embed/gpu/nvidia-geforce-rtx-4090/"
  width="440" height="320"
  frameborder="0"
  title="NVIDIA GeForce RTX 4090 — GPU Ark">
</iframe>

Replace nvidia-geforce-rtx-4090 with the slug of the desired GPU (see API).

GPU Comparison

Side-by-side comparison table of two GPUs. Best values are highlighted.

Embed code:

<iframe
  src="https://gpuark.com/en/embed/compare/nvidia-geforce-rtx-4090-vs-nvidia-geforce-rtx-3090/"
  width="640" height="460"
  frameborder="0"
  title="RTX 4090 vs RTX 3090 — GPU Ark">
</iframe>

GPU Ranking by GPI

Top GPUs by GPU Performance Index. Filter by vendor and set the count.

Embed code:

<iframe
  src="https://gpuark.com/en/embed/top/?vendor=nvidia&limit=10"
  width="540" height="580"
  frameborder="0"
  title="Top GPU по GPI — GPU Ark">
</iframe>

Parameters: vendor (nvidia, amd, intel), limit (1–50, default 10).

Terms of Use

The API is free for any use — commercial and non-commercial.

The only requirement: when using GPU Ark data, include a link to the source:

Data: <a href="https://gpuark.com">GPU Ark</a>

This helps us continue to maintain and develop the free service.

We reserve the right to restrict access in case of abuse (DDoS, mass scraping without attribution).

About the Data

The database contains 12,013+ GPUs with detailed specifications:

VRAM, memory type, bus width, bandwidth
CUDA cores, TMU, ROP, Tensor Cores, RT Cores
Performance: FP16, FP32, FP64, BF16, TF32, INT8 (TFLOPS)
NVENC / NVDEC: generation, codecs (H.264, H.265, AV1, VP9)
API: CUDA, Vulkan, OpenCL, OpenGL, DirectX
GPI — GPU Performance Index (proprietary metric)

Data is automatically updated from public sources.

Contact

We welcome suggestions here: t.me/maxvyaznikov