Skip to content
Kinesis

App Gallery · open models and apps, ready to run

Pick a model or an app. Your Kinesis Grid finds the GPUs and keeps it running.

Open LLMs such as gpt-oss, Gemma, Qwen and DeepSeek, each with its own private chat, any model from the Ollama library, and more than 20 apps from JupyterLab to Grafana. Choose where it runs and the most you’ll pay, and it comes up on its own HTTPS address in a few clicks.

Get startedSee what’s inside

Small models can try out on CPU. Larger models need a GPU in your Grid.

What’s inside

Open models, and the tools around them.

Each template is a ready-made app: the image, ports, settings and hardware it needs. Launch it as it is, or change anything first. Not only LLMs: notebooks, dev tools, dashboards and business apps are one click away too.

Open LLMs, each with its own private chat

Open WebUI, a ChatGPT-style chat, on Ollama: set up for you, on GPUs in your Grid.

  • App Gallery card for choosing any Ollama model

    Choose Custom Model

    Any model from the Ollama library

    Search by name, pick a size up to 130B parameters, and Kinesis sizes the GPU memory for it. Llama, Mistral, Phi and the rest of the library.

  • App Gallery card for OpenAI gpt-oss

    gpt-oss:120b

    OpenAI gpt-oss

    1 GPU40 GiB VRAM
  • App Gallery card for Google Gemma 4

    gemma4:31b

    Google Gemma 4

    1 GPU48 GiB VRAM
  • App Gallery card for Qwen 3.6

    qwen3.6:27b

    Qwen 3.6

    1 GPU48 GiB VRAM
  • App Gallery card for DeepSeek-R1

    deepseek-r1:32b

    DeepSeek-R1

    1 GPU64 GiB VRAM
  • App Gallery card for Google Gemma 4

    gemma4:e4b

    Google Gemma 4

    1 GPU16 GiB VRAMCPU trial
  • App Gallery card for Qwen 3.5

    qwen3.5:9b

    Qwen 3.5

    1 GPU16 GiB VRAMCPU trial

Also in the gallery: Microsoft BitNet on llama.cpp, which needs no GPU. Models marked CPU trial can also start on serverless CPU; larger ones run on a GPU in your Grid.

More than 20 apps around them

Popular open-source apps, set up to run on your Grid. Most need no GPU at all.

  • Notebooks and dev tools

    • JupyterLabNotebooks for data science and ML
    • code-serverVS Code in your browser
    • GiteaYour own Git server

  • Dashboards and automation

    • GrafanaDashboards and alerts
    • Uptime KumaUptime checks and status pages
    • n8nWorkflow automation
    • changedetection.ioAlerts when a web page changes

  • Business apps

    • AppsmithInternal tools, built visually
    • NocoDBA database that works like a spreadsheet
    • KanboardKanban boards for your projects
    • draw.ioDiagrams in the browser

  • AI building blocks

    • Whisper ASRSpeech to text as a web service
    • QdrantA vector database for RAG
    • MeilisearchFast search over your own data
    • LibreTranslateTranslation you host
    • SearXNGPrivate web search

  • Files and documents

    • MinIOS3-compatible object storage
    • FileBrowserYour files, in the browser
    • Stirling PDFEdit, merge and convert PDFs
    • GotenbergDocuments to PDF through an API

How it works

From the gallery to a live app in four steps.

  1. 1

    Pick a template

    Choose a model or an app, or name any Ollama model. Each template shows the GPUs, memory and disk it needs.

  2. 2

    Choose where it runs

    Apps without a GPU can run on serverless CPU. GPU apps run on your Kinesis Grid: Kinesis capacity, your cloud accounts or your own servers, under the max price per GPU-hour you set.

  3. 3

    Launch, or customize first

    Quick Launch starts it as it is. Customize First opens every setting: ports, environment variables, hardware and location.

  4. 4

    Use it on its own HTTPS address

    Your Grid finds the GPUs and starts the app. When it is ready, it answers on its own HTTPS address, with logs and metrics in the portal.

New appfrom the App Gallery

gemma4:31b

Open WebUI with Ollama

Where it runs

Your Kinesis Grid1 GPU48 GiB VRAMYour max price per GPU-hour
Quick LaunchCustomize First
  1. Your Grid found a GPU that fits
  2. Model downloaded, chat ready
  3. HTTPS address live

Illustrative

Up and running

In our benchmark, Kinesis was 4.1× faster than AWS and 6.3× faster than Azure.

We timed one job end to end, sign-up included: a production HTTPS model endpoint (vLLM serving Gemma 4 31B FP8) on one NVIDIA H100.

  • Kinesisabout 27 min · 93 steps
  • AWSabout 1.85 h · 337 steps
  • Azureabout 2.8 h · 483 steps

HTTPS on a managed domain comes built in: 0 steps for the gateway and TLS on Kinesis, against 119 on AWS and 219 on Azure.

Kinesis Builder Time-to-Value Benchmark, May 2026. Wall-clock time from sign-up to a live endpoint; a step is one click, command or form entry.

Private by design

Your prompts stay in your Grid.

A model from the App Gallery answers from GPUs in your own Kinesis Grid, not from a model vendor’s API. Your prompts never go to a model vendor, and you pay for GPU hours, not tokens. Keep the frontier APIs for the hardest problems.

  • Switch models when a better one comes out: Gemma 4, Qwen 3.5 and Qwen 3.6 joined the gallery as they were released.
  • Your Grid never places work on community hardware unless you choose Kinesis Batches.
  • Grids are private to your organization unless your admins share them.
  • Limit an app to your own IP addresses when it should not be public.
Open WebUIgemma4:31b
HTTPS · in your Kinesis Grid

Summarize this supplier contract in five bullets.

Here are the key terms: a 24-month term, net-60 payment, and…

Answered by a GPU in your Kinesis Grid. Nothing went to a model vendor.

Illustrative

Comes with every app

Everything around the app is already there.

  • Its own HTTPS address

    Certificates come automatically, and you can add your own domain.

  • Logs, metrics and a terminal

    Follow the startup, watch the usage and open a shell in your browser.

  • Kept running by your Grid

    If a machine fails or a spot GPU leaves, your Grid moves the app to another machine that fits.

  • Settings you can change

    Ports, environment variables, hardware and location, before you launch or later.

  • Shared with your team

    Apps live in your organization’s projects. Add teammates as viewers or developers.

  • Ready for your agents

    The REST API and MCP server list every template, so Claude, ChatGPT or Cursor can launch one for you.

Questions

What people ask about the App Gallery.

What is in the App Gallery?
Open LLMs with a chat interface (gpt-oss, Gemma 4, Qwen 3.5 and 3.6, DeepSeek-R1 and Microsoft BitNet), a picker for any model in the Ollama library, and more than 20 apps: notebooks, dev tools, dashboards, automation, search, storage and document tools. Each template is a complete app configuration you can launch as it is or change first.
Do I need GPUs of my own?
No. Your Kinesis Grid can use Kinesis capacity, the cloud accounts you connect with an API key, or servers you own, in any mix. Small models, up to about 9B parameters, can also try out on serverless CPU.
Can I bring my own app or repo?
Yes. Next to the gallery, you can deploy any container image, or build from a GitHub repository or a Dockerfile. Your app runs on the same Grid, with the same HTTPS address, logs and metrics.
What does it cost?
The App Gallery itself costs nothing. You pay for the compute your app runs on: GPUs by the GPU-hour, never per token, and a spot GPU never costs more than the max price you set. For compute you bring, Kinesis charges a small Grid fee instead. The details are on our pricing page.
Who can reach my app?
Anyone with its HTTPS address, so treat the address like a key. The chat templates start without a sign-in, and some apps ship with a default password: set your own passwords, or limit the app to your IP addresses, before you share it.
What happens when a machine fails?
Your Kinesis Grid replaces it. It finds another machine that fits your app and your max price, starts the app there, and tells you in your notifications.

Your next model is a few clicks away.

Create an account, pick a model or an app, and your Kinesis Grid finds the GPUs and keeps it running.

Get startedSee pricing