TL;DR
Nemotron 3 Ultra 550B produced valid structured actions 77.5% of the time in the August 8, 2026 next-action replay. Its judge mean was 57.1 and historical billing was $35.28 per 1,000 decisions; production median timing was unavailable. The published Luna Pro arm scored 91.7 with 100% valid output at $25.05 per 1,000 decisions. This measures individual agent decisions, not an entire trading run.
- Exact tested ID
- nvidia/nemotron-3-ultra-550b
- Published studies
- 1
- Provider
- nvidia
Next-action execution model bakeoff: compare the tested outputs
These exact August 8, 2026 percentages measure output validity on 69 decisions sampled three times each. Labels are rounded; the full table retains judge scores and billing. Production timing remains unavailable for Nemotron.
Show the table · 2 rows, 5 columns
Scroll sideways to compare all columns. Select a heading to sort.
| nvidia/nemotron-3-ultra-550b | 57.1 | 77.5% | $35.28 | n/a |
| openai/gpt-5.6-luna-pro | 91.7 | 100.0% | $25.05 | 9.9s |
No matching rows. Clear the filter to see all records.
23-model campaign; 69 frozen decisions × 3 samples; 12 published model rows. Published 2026-08-08. Values retain their original units and grading scale.
nvidia/nemotron-3-ultra-550b was tested on 2026-08-08. 23-model campaign; 69 frozen decisions × 3 samples; 12 published model rows. Its row and the other published arm use the same task and measurement units.
What this version helps you choose
Output validity was a material issue in this next-action replay: 77.5% of structured actions met the contract. Production median timing is unavailable.
Choose the next allowed agent action: follow the published task
A well-formed action can still be the wrong decision. Keep output validity and judged decision quality separate.
Published task summary
The existing plan, conversation and results returned by tools.
Choose one next allowed action, or finish with an answer when the evidence is sufficient.
Read what came back
A structured action. The replay records the proposed action without executing its tools.
Replay 69 saved decisions three times each. Check the required output structure separately from the judge score.
Next-action execution model bakeoff
Inspect the original Nemotron metrics
For the next allowed action, Nemotron 3 Ultra 550B recorded a judge mean of 57.1 and produced valid structured output 77.5% of the time. Historical billing was $35.28 per 1,000 decisions; production median response time was n/a. Decision quality, readable output and operating cost answer different questions.
| Published metric | Value |
|---|---|
| Mean score | 57.1 |
| Schema valid | 77.5% |
| $ / 1k decisions | $35.28 |
| Prod p50 | n/a |
No matching rows. Clear the filter to see all records.
Next-action execution model bakeoff: read these measurements
Read the complete rubric, sample and billing notes
23-model campaign; 69 frozen decisions × 3 samples; 12 published model rows
Choose one next action in an unfinished trading-agent workflow. This replay measures individual decisions, not a fresh complete trading run.
Scoring
Mean action score comes from the judge. Schema validity is the share of actions matching the required output structure; a valid structure can still contain the wrong action. Switching judges reduced the Luna-versus-DeepSeek score gap from 13.9 to 1.5 points.
Cost units
Cost uses historical provider billing per 1,000 decisions. Reused prefixes earned cache discounts; the cached-token split was not retained.
Time units
Prod p50 is median production response latency, separate from replay quality and complete agent wall time.
No comments yet.