Any model fits the GPU you already have. It runs out of memory so you don't — 100% local, never a rented GPU. Pick your platform — once it's running, a local web UI opens in your browser.
vrampilot.exe; on the blue Windows protected your PC box click More info → Run anyway.chmod +x vrampilot-0.2.0-linux-x64.bin — then run ./vrampilot-0.2.0-linux-x64.bin.chmod +x vrampilot-0.2.0-macos-arm64.bin — then run ./vrampilot-0.2.0-macos-arm64.bin. It's a command-line binary, so the notarization ticket is checked online; if macOS blocks it the first time, clear the download flag with xattr -d com.apple.quarantine vrampilot-0.2.0-macos-arm64.bin or allow it in System Settings → Privacy & Security → Open Anyway.Every build has a published SHA-256. On your machine run sha256sum (or shasum -a 256) and match it against the list.
Open-source benchmarks & raw runs: github.com/zmlabs-ai/benchmarks