GPU Acceleration

GPU acceleration enables GPU-accelerated rendering in Kernel browsers, providing enhanced performance for graphics-intensive workloads.

GPU acceleration is only available for headful browsers and does not support standby mode. GPU browsers have separate pricing from headless or headful browsers.

Enable GPU acceleration

Set the gpu parameter to true when creating a browser:

import Kernel from '@onkernel/sdk';

const kernel = new Kernel();

const kernelBrowser = await kernel.browsers.create({
  gpu: true
});

from kernel import Kernel

kernel = Kernel()

kernel_browser = kernel.browsers.create(
    gpu=True
)

package main

import (
	"context"

	"github.com/kernel/kernel-go-sdk"
)

func main() {
	ctx := context.Background()
	client := kernel.NewClient()

	kernelBrowser, err := client.Browsers.New(ctx, kernel.BrowserNewParams{
		GPU: kernel.Bool(true),
	})
	if err != nil {
		panic(err)
	}
	_ = kernelBrowser
}

You can also enable GPU acceleration in the dashboard when deploying a browser under Advanced Configuration.

Large GPU browsers

A standard GPU browser runs with 4 vCPUs, 6 GiB of memory, and a 2 GiB vGPU frame buffer. For graphics-heavy pages, request the large GPU browser by setting memory to 12GiB. It doubles CPU, memory, and frame buffer: 8 vCPUs, 12 GiB of memory, and a 4 GiB frame buffer.

import Kernel from '@onkernel/sdk';

const kernel = new Kernel();

const kernelBrowser = await kernel.browsers.create({
  gpu: true,
  memory: '12GiB',
});

from kernel import Kernel

kernel = Kernel()

kernel_browser = kernel.browsers.create(
    gpu=True,
    memory="12GiB",
)

package main

import (
	"context"

	"github.com/kernel/kernel-go-sdk"
)

func main() {
	ctx := context.Background()
	client := kernel.NewClient()

	kernelBrowser, err := client.Browsers.New(ctx, kernel.BrowserNewParams{
		GPU:    kernel.Bool(true),
		Memory: "12GiB",
	})
	if err != nil {
		panic(err)
	}
	_ = kernelBrowser
}

12GiB is the only memory value accepted together with gpu: true; omit memory to get the standard GPU browser. The memory field on the browser response reports the allocation you received. Large GPU browsers come from their own capacity. When none is available, creation fails with HTTP 529 and the no_gpu_browsers_available code rather than falling back to a standard GPU browser, so retry later or create a standard GPU browser explicitly.

Use cases

GPU acceleration is useful for:

High performance live view streaming
Rendering complex 3D graphics or WebGL content
Video processing and playback
Canvas-heavy applications

GPU acceleration speeds up page rendering, not screenshot capture. CDP’s Page.captureScreenshot latency is unaffected by GPU. Computer controls screenshots see a small speedup with GPU enabled.

Availability

GPU acceleration is available on Start-Up and Enterprise plans.

Overview

Working with your browser

Integrations

deploying your agent

Agent Skills

Community

Info

GPU Acceleration

Enable GPU acceleration

Large GPU browsers

Use cases

Availability

​Enable GPU acceleration

​Large GPU browsers

​Use cases

​Availability

Enable GPU acceleration

Large GPU browsers

Use cases

Availability