Hermes HQ Subscribe
Models

Run Hermes Agent on NVIDIA's Nemotron 3 Ultra model with a free key

What it does

Nemotron 3 Ultra, a 550B model, can run Hermes Agent on a free NVIDIA Build key. The key is added to ~/.hermes/.env as NVIDIA_API_KEY, a single test chat tries the model, and hermes model with NVIDIA NIM keeps it.

In a test on a small project with a failing test, Hermes Agent was asked to fix the code without touching the tests, and three runs produced three fixes.

Why it's worth it

Nemotron 3 Ultra, a 550B model, can run Hermes Agent on a free NVIDIA Build key, so you can try a large model without a paid key.

Get a failing test fixed with Nemotron 3 Ultra, without touching the tests.

Step by step

Paste it into Hermes Agent and swap the highlighted parts for your own details.

Prompt

Before you start A free NVIDIA Build API key, saved as NVIDIA_API_KEY in ~/.hermes/.env.

Get Hermes Agent to:

  1. Check my ~/.hermes/.env file for an NVIDIA_API_KEY line, and if it is missing or empty, stop and tell me to add my free NVIDIA Build key there as NVIDIA_API_KEY.
  2. Send one test chat to nvidia/nemotron-3-ultra-550b-a55b through NVIDIA NIM and confirm it replies before going further.
  3. Ask me to approve before changing my default model, and wait for my yes.
  4. Once I approve, keep NVIDIA NIM with nvidia/nemotron-3-ultra-550b-a55b as my default model.
  5. In [project folder], find the failing test [test name] and read the code it covers without editing anything yet.
  6. Fix the code so [test name] passes, leave every test file unchanged, and run the test suite to confirm the result.
  7. Repeat the fix three times and list each run's result, noting any run that changed a test file.
  8. Report back with the model used, whether each run passed, and which files I should review.

Expected resultThree runs on a failing test in a small project each produced a fix.

Related use cases

See all
Prompt