Submit your results

Run our free benchmark on your PC and share the numbers. It measures how fast local AI models run on your hardware with Ollama. Submissions are reviewed before they appear on the Benchmarks page, labelled Community.

Download the benchmark kit (.zip, 10 KB)

Privacy: we send only the benchmark numbers, model names, your rig label, OS version, CPU and GPU model names. No computer name, user name or file paths. Your IP address is used only for spam rate-limiting, stored as a one-way hash. You see exactly what is sent and must type y before anything leaves your PC.

Windows: step by step

About 15 minutes of setup, then 10 to 40 minutes of waiting. Plug laptops into power.

Install Ollama. Download it from ollama.com/download, run the installer, and leave it running. You will see a small llama icon near the clock.
Install Python 3. Download it from python.org/downloads and run the installer. On the first screen, tick "Add python.exe to PATH", then click "Install Now".
Unzip the kit. Find fitgpu-bench-kit.zip in your Downloads folder, right-click it, choose Extract All…, then click Extract. A folder opens containing another folder, fitgpu-bench. Double-click it to open it.
Open PowerShell in that folder. Click once on the address bar at the top of the folder window (where it shows the folder path), type powershell, and press Enter. A blue or black text window opens. That window is PowerShell, already "inside" the kit folder.
Allow the script to run (this window only). Copy this line, right-click inside the PowerShell window to paste it, and press Enter:
```
Set-ExecutionPolicy -Scope Process Bypass
```
If it asks to confirm, type Y and press Enter. This only affects this one window and is undone when you close it.
Run the benchmark. Paste the line below, but first change the text in quotes to describe your PC (graphics card, memory). Don't put your name in it. Then press Enter:
```
.\run_bench.ps1 -Rig "RTX 4070 12GB, 32GB RAM" -Submit
```
It downloads the test models (several GB) and then runs them. Don't use the PC for games or video while it runs.
Review and send. At the end it prints exactly what will be sent and asks Send these results? [y/N]. Type y and press Enter to share, or just press Enter to keep them private. Either way, a results-….json file is saved in the kit folder.

Something went wrong? "py is not recognized" means Python isn't installed or "Add to PATH" wasn't ticked: re-run the Python installer and choose "Modify". "Could not connect" means Ollama isn't running: start it from the Start menu. You can also skip -Submit and upload the results file below.

Linux / macOS: python3 ollama_bench.py --rig "your rig" --models llama3.1:8b qwen3:8b --submit

Or upload a results file

Already ran it? Choose the results-….json file from the kit folder. You'll see what will be sent before you confirm.