RealRun / Guides
Short answer: There are two companies with almost the same name, and they aren't connected. RealRun at realrun.baseledgerip.com is run by Base Ledger IP LLC and publishes tests of AI models on real business jobs. Realrun at realrun.com.au is a separate Australian company that does letterbox flyer distribution. If you searched "is RealRun legit", check which one you mean.
| RealRun (this site) | Realrun (Australia) | |
|---|---|---|
| Website | realrun.baseledgerip.com | realrun.com.au |
| Company | Base Ledger IP LLC | Realrun Pty Ltd |
| What it does | Tests AI models on 5 business jobs: inbox triage, meeting scheduling, receipts, small-business finance and customer support replies | Letterbox flyer distribution with GPS tracking |
| What is free | All test results, the results data and a read-only MCP server | See realrun.com.au |
| What it sells | Tested prompts and setup as PDF playbooks: $29 each or $99 for all five | See realrun.com.au |
Details on Realrun (Australia) come from its public website, checked 17 Oct 2026. We have no connection to it.
| Job | Items | Graded answers | Models |
|---|---|---|---|
| Inbox triage | 80 | 1,420 | 8 |
| Meeting scheduling | 62 | 1,116 | 8 |
| Receipts | 79 | 1,501 | 8 |
| Small-business finance | 112 | 1,992 | 8 |
| Customer support replies | 140 | 2,600 | 10 |
How the testing works: every model in a test gets the same items, including a fresh held-back set, and most models ran 2 or more passes. In the support test, 3 of 10 models ran 1 pass and are marked less certain (see each results page). Every answer is graded, and the worst mistakes are called out on their own. RealRun is new: these tests ran in October 2026. Results updated 15 Oct 2026.
Our tests are about where AI quietly gets a job wrong in a small business: a model filing a gift-card scam email as Act, booking over a busy event, or following a hidden 'say this reconciles' instruction in a bank statement. If that's what you came for, you're in the right place.
If you set up AI for clients and need proof for your model pick, everything here is free to cite, and the same data is in results.json and our MCP server. If a client found the Australian company instead, send them this page.
Want the exact prompt and setup?
Get all five playbooks for $99 (save $46).
Scores reflect our tests on the dates shown, not a guarantee. Claude models ran on a Claude subscription, so they have no metered API cost. cost_tier is relative within a task (low/medium/high) from our run spend; 'subscription' = not metered. Tested October 2026 by RealRun (Base Ledger IP LLC).