How to Use Nemotron 3 Ultra in Cursor
Nemotron 3 Ultra handles complex coding, research synthesis, and planning across long conversations. Its 550B total and 55B active parameters distinguish it from Nemotron 3 Super’s 120B total and 12B active parameters.
Prompts to try with Nemotron 3 Ultra in Cursor
Copy an example into Cursor, or ask in your own words.
Use Nemotron 3 Ultra to investigate the failing tests in this project. Trace the relevant code, propose the smallest justified patch, and add a regression test that fails before the fix.
Use Nemotron 3 Ultra to review this repository's authentication flow. Map the trust boundaries, cite the relevant code, and rank findings by demonstrated impact.
Use Nemotron 3 Ultra to plan a migration of this project's database access layer. Identify dependencies, divide the work into testable stages, and define rollback criteria.
Use Nemotron 3 Ultra to compare the implementation with the specification in this project. Return a table of mismatches with code references and a proposed verification step for each.
Use Nemotron 3 Ultra to review the numerical calculations in this repository. Check assumptions and edge cases, then add tests for any errors you can reproduce.
How to set up Nemotron 3 Ultra in Cursor
Pick the way that suits you. Each takes about two minutes.
Use Spiral's MCP server
- Add the hosted server
Use Add to Cursor on the console's Agents & MCP page, or add this to .cursor/mcp.json (one project) or ~/.cursor/mcp.json (every project):
{ "mcpServers": { "spiral": { "url": "https://api.spiral.ai/mcp" } } } - Sign in
Open Cursor Settings › MCP, click Connect next to spiral, sign in to Spiral and click Authorize. No API key. Spiral then shows as connected, with its tools listed.
Use the Spiral skill
- Send Cursor this message
Your agent reads Spiral's skill and sets itself up.
set up https://spiral.ai/SKILL.md
- Sign in or paste a key
Approve Spiral in the browser, or give the agent a key from the console.
Use the Spiral CLI
- Install and sign in
Installs the Spiral CLI, signs you in and adds its skill to your coding agents.
npx -y spiral-cli@latest init
Call the Spiral API
- Create an API key
In the console, under API keys.
- Run Nemotron 3 Ultra
One request; the result comes back by polling or a webhook.
curl https://api.spiral.ai/v1/runs \ -H "Authorization: Bearer $SPIRAL_API_KEY" \ -H "Content-Type: application/json" \ -H "Idempotency-Key: $(uuidgen)" \ -d '{ "provider": "openrouter", "model": "nvidia/nemotron-3-ultra-550b-a55b", "input": { "messages": [ { "role": "system" } ] } }'
Then paste one of the prompts above: Cursor finds Nemotron 3 Ultra on Spiral, shows the price and runs it.
Nemotron 3 Ultra providers and prices
Spiral runs Nemotron 3 Ultra through OpenRouter. Open a card for today's price, with every fee included.
Up to $0.011738 per run at the default inputs. $0.65 per 1M input tokens, $2.86 per 1M output tokens.
Settings you can ask for
Ask Cursor for any of these in your prompt.
| Setting | Allowed values | Default | Description |
|---|---|---|---|
| Seed | a whole number | ||
| Min p | 0 to 1 | ||
| Top k | a whole number from 0 | ||
| Top p | 0 to 1 | ||
| Max tokens | 1 to 182520 | The most output tokens to generate. | |
| Temperature | 0 to 2 | ||
| Presence penalty | -2 to 2 | ||
| Reasoning effort | none, medium, or high | high | How much the model reasons. |
| Frequency penalty | -2 to 2 | ||
| Repetition penalty | 0 to 2 |
Every API field of NVIDIA: Nemotron 3 Ultra is on its tool page: open its card above.
What can you do with Nemotron 3 Ultra in Cursor?
Turn a large codebase and its failure evidence into a concrete repair or migration plan with Nemotron 3 Ultra. Give the agent acceptance criteria so its analysis ends in changes you can test.
Plan a codebase migration
Compare the current architecture with a proposed replacement, trace dependencies, and draft migration stages with tests and rollback criteria.
Investigate a software defect
Analyze code and failure logs, propose a cause, and produce a patch with a regression test.
Reconcile research findings
Compare conflicting sources, separate agreement from disagreement, and write a brief that ties conclusions to evidence.
Answer questions across long documents
Analyze supplied manuals, policies, or reports and return answers with supporting passages and unresolved questions.
Review multilingual support cases
Compare English, Spanish, French, or other supported-language conversations against supplied policies and draft responses in the customer's language.
Check mathematical and scientific arguments
Examine supplied assumptions, calculations, and code; identify gaps and propose checks that could confirm or reject the result.
Turn operational evidence into an action plan
Combine tool results and incident timelines into ranked hypotheses, verification steps, and a recovery plan.
Extract structured findings
Convert supplied reports into consistent records of findings, evidence, and missing information for downstream review.
More on Nemotron 3 Ultra and Cursor
Overview
Nemotron 3 Ultra in other agents
Questions about Nemotron 3 Ultra in Cursor
Can Cursor use Nemotron 3 Ultra?
Yes. Add Spiral to Cursor as an MCP server, and Cursor can find and run Nemotron 3 Ultra like any built-in tool.
How much does Nemotron 3 Ultra cost in Cursor?
Up to $0.012808 per run at the default inputs. $0.78 per 1M input tokens, $3.12 per 1M output tokens. You pay from your Spiral balance as you go, with no subscription, and a failed run is not charged. Your Cursor plan is separate.
Do I need an OpenRouter account to use Nemotron 3 Ultra in Cursor?
No. Spiral runs Nemotron 3 Ultra on its own OpenRouter account. You sign in to Spiral once and never paste a key.
How long does it take to set up Nemotron 3 Ultra in Cursor?
About two minutes: add the hosted server, then sign in.
Does Cursor wait while Nemotron 3 Ultra runs?
Nemotron 3 Ultra runs in the background. Cursor starts the run, checks on it and gives you the text, json when it's ready.
What is Nemotron 3 Ultra good at?
Use it for tasks that combine planning, code generation, research, and verification. NVIDIA specifically highlights maintaining architectural decisions across coding sessions and reconciling contradictory research evidence.
How much context can Nemotron 3 Ultra handle?
NVIDIA documents up to 1 million tokens of context. For document analysis, ask for source labels and supporting passages so you can check the conclusions.
Which languages does Nemotron 3 Ultra support?
NVIDIA lists English, French, Spanish, Italian, German, Japanese, Hindi, Korean, Brazilian Portuguese, and Chinese. State the desired answer language explicitly when your source material mixes languages.
What is the tradeoff with shorter reasoning?
Nemotron 3 Ultra was trained for regular, medium-effort, and reasoning-off modes. NVIDIA reports approximately 7% lower accuracy for medium-effort reasoning in its Artificial Analysis comparison, so test reduced reasoning on representative tasks before using it for difficult decisions.
Does Nemotron 3 Ultra know recent events?
NVIDIA lists May 2026 as its post-training data cutoff. Supply current evidence for newer events and ask it to distinguish source-backed findings from assumptions.
Start using Nemotron 3 Ultra in Cursor
One Spiral account for every tool your agent needs. No subscription; you pay per run.
Paste this into Cursor:
set up https://spiral.ai/SKILL.md


