resyn

Run the benchmark

Your model answers the size you pick, graded and timed by the site.

Racing uwu: Gemma-4-31B 22 of 44 tasks · 13.1 tok/s. Run the same model here and the two go side by side.

The app

Running Ollama, LM Studio or llama.cpp with a model loaded? Press Connect, and choose Allow if your browser asks.

First time: let this site talk to Ollama Ollama only answers pages it trusts. Tell it once to trust https://resyn.cloud, then start it again. Pick your system:
Mac
  1. Quit Ollama: click its icon in the menu bar and choose Quit Ollama. Closing its window is not enough.
  2. Open Terminal: press Command-Space, type Terminal and press Return.
  3. Copy this line, paste it into the Terminal window and press Return. It prints nothing when it works.

    launchctl setenv OLLAMA_ORIGINS https://resyn.cloud

  4. Open Ollama again, then press Find my models below. If it finds your models, this step is done.

This lasts until the Mac restarts. If the page stops reaching Ollama after a restart, run the line again, then quit and reopen Ollama.

Windows
  1. Quit Ollama: right-click its icon by the clock in the taskbar and choose Quit Ollama. If you cannot see it, click the small arrow by the clock first.
  2. Press the Windows key, type environment variables and open "Edit environment variables for your account".
  3. Under "User variables" click New. Variable name:

    OLLAMA_ORIGINS

    Variable value:

    https://resyn.cloud

    Click OK, then OK again.
  4. Open Ollama from the Start menu, then press Find my models below. If it finds your models, this step is done.
Linux
  1. In a terminal, run this. An editor opens.

    sudo systemctl edit ollama

  2. Type these two lines on the empty line between the two comment blocks (the lines starting with #). Then save and close: in nano, press Ctrl+O, Return, then Ctrl+X.

    [Service] Environment="OLLAMA_ORIGINS=https://resyn.cloud"

  3. Run this, then press Find my models below. If it finds your models, this step is done.

    sudo systemctl restart ollama

If you start Ollama by hand instead of as a service, run OLLAMA_ORIGINS=https://resyn.cloud ollama serve.

New to this? Start with a small model
  1. Install Ollama, a free app that runs AI models on your computer, and open it.
  2. Open a terminal (on a Mac, the Terminal app; on Windows, PowerShell) and paste this line. It fetches a tiny model, about 270 MB:

    ollama pull smollm2:135m

  3. Press Connect above. If Ollama refuses this page, do the one-time step that appears, then press Connect again. Pick smollm2:135m and run.

SmolLM2 has 135 million parameters, so a first run shows how the benchmark works. Try a bigger model next. The number in a model's name (4B, 8B) is its size in billions of parameters, and the home page suggests one that fits.

Choose the app yourself, or use Claude, GPT and others online
App

For Claude, GPT or Gemini, pick OpenRouter: one key serves them all.

Leave this as it is unless you changed the app's port.

Or let a coding agent do it

Using Claude Code, Codex CLI, Gemini CLI or Cursor on this computer? Paste this to it: it runs the model and posts the result.

Run the resyn benchmark on the model I have running on this computer and give me the link to the result. If more than one app or model is running, ask me which one. Read https://resyn.cloud/llms.txt first and do what it says.

More options: another app, or posting under your name.