stealthmodels.

MODEL BENCHMARKS

Nemotron 3 Ultra

Combines Mamba, attention and mixture-of-experts layers in a 550B-parameter model, with 55B active and multi-token prediction for faster serving.

BENCHMARK BY BENCHMARK

Nemotron 3 Ultra benchmark scores

StealthMark
16.7/100Overall rank 55/58
General Intelligence
12.2/100
Visual Intelligence
26.5/100

Nemotron 3 UltraOther modelsMedian

Reasoning efficiency

How many reasoning tokens a model needs for the quality it delivers.

Single-step reasoningHow much it reasons to understand a taskRank 45/56
Multi-step reasoningHow much it keeps reasoning as a task goes onRank 35/54

Cost efficiency

How much Performance you get for your money.

Rank 36/52

DATASHEET

Nemotron 3 Ultra specs and pricing

A good fit forSelf-hosted reasoning, tool use and retrieval over long documents on multi-GPU NVIDIA infrastructure.

Nemotron 3 UltraOther modelsMedian

PriceInput / output per 1M tokens
$0.60 / $2.40Cheaper than 26 of 46 othersFrom $0.90 / $1.70 with other providers
Context windowTokens in one request
1,000,000Larger than 7 of 49 others
Output speedTokens per second, our measurement
139 tokens/sFaster than 27 of 51 others
Developer
NVIDIA
Released
4 Jun 2026
Family
Nemotron 3 Ultra
Open model
Yes
Local model
No
Origin
United States
Parameters
550B
Active parameters
55B
Input → output
Text → text
API model ID
nvidia/nemotron-3-ultra-550b-a55b
Architecture
LatentMoE: Mamba-2 + MoE + Attention hybrid with Multi-Token Prediction

Self-host Nemotron 3 Ultra

Hardware
Multi-GPU datacenterEstimated

Official Hugging Face weightsOfficial NVFP4 weights

All prices and providers3

OpenRouter via NVIDIA Build; input 0.5 USD/1M; output 2.2/1M; Partner endpoint price displayed by NVIDIA Build; not a universal model price.

  • OpenRouterObserved 2026-10-23
    Input $0.6Output $2.4Cached read $0.12

    per 1M tokens

    nvidia/nemotron-3-ultra-550b-a55b
  • OpenRouter via NVIDIA BuildObserved 2026-10-06
    Input $0.5Output $2.2

    per 1M tokens

    Available

    nvidia/nemotron-3-ultra-550b-a55b
  • Input $0.9Output $1.7

    per 1M tokens

    Available

    nvidia/nemotron-3-ultra-550b-a55b
Measurements and sources6

GAME BENCHMARKS

Watch Nemotron 3 Ultra play

Foresight

Anticipating threats and planning ahead

16.9/100Rank 48/54
Survived 0 of 3
Foresight: complete playing field
Level 38 ptsTurn 17 / 20

Points per level

  1. Level 1Second Thoughts4 ptsBest 27Out of lives
  2. Level 2Pressure Pot4 ptsBest 13Out of lives
  3. Level 3Hide or Flight8 ptsBest 19Out of lives

Collect coins while avoiding ghosts. Ghosts move before the player each turn, and collisions cost a life. 3 levels in sequence, up to 66 turns in total. Earlier moves and outcomes carry forward, testing learning across levels.

Latch

Cooperative planning and learning from experience

5.0/100Rank 49/55
4 of 5 played
Latch: complete playing field
Level 413.8 ptsTurn 21 / 40

Points per level

  1. Level 1Hold the door5.2 ptsBest 5.2Cleared
  2. Level 2Crossfeed3.2 ptsBest 10.2Finished early
  3. Level 3Relay vault7.4 ptsBest 19.2Finished early
  4. Level 4Passing bay13.8 ptsBest 18.2Finished early
  5. Level 5Second thoughtsNot played

Coordinate two robots to push numbered cores onto matching sockets. Pressure plates hold doors open; moving either robot uses one turn. 5 levels in sequence, up to 152 turns in total. Earlier moves and outcomes carry forward, testing learning across levels.

Nemotron 3 Ultra benchmark results ↗

StealthMark score

StealthMark measures how well a model’s abilities carry over to original, demanding tasks it could not have trained for: physical and spatial reasoning, planning across many steps, and precise visual work. How StealthMark works explains every benchmark, what it tests and why it is hard.

General Intelligence

The ability to reason, plan, solve problems and understand how things relate and behave. A high score means it thinks hard problems through, from careful analysis to plans that span many steps.

Visual Intelligence

The ability to understand visual intent and turn ideas, descriptions and plans into coherent, effective visuals. A high score means design ideas come out the way you meant them, in illustrations, icons, diagrams or interface visuals.

Overall score

Overall shows how well a model’s abilities carry over to original, demanding tasks, weighting every benchmark into one score. Estimated means the total includes an estimate for one missing scene.

Judging

Scores reflect how well each result meets the benchmark requirements. Game scores come from the recorded play: the points earned on each level and how each level ended. How StealthMark keeps scores honest.

Games

Each game is a sequence of levels, and earlier moves carry into later levels. A level that was not played is shown as not played and never counts as zero. Play every level of Latch or Foresight yourself to get a score calculated like the models’.

Reasoning efficiency

Good results with little thinking score high. On each task, a model’s result and reasoning tokens are compared with every other model and combined with the Balanced Integration Score (Liesefeld & Janczyk, 2019); 50 is average. Single step shows how much reasoning a model needs to understand and solve one problem, multi-step how efficiently it keeps working on a task by building on its earlier reasoning, as in agentic workflows. Not part of StealthMark.

Cost efficiency

How much StealthMark a model delivers for its money, in StealthMark points per dollar: its overall score divided by what one benchmark task costs it on average. The cost is what our runs were charged, or their tokens at the model’s list price; benchmarks a model has not run yet are estimated from its other costs. Like StealthMark, it is absolute and compares with no other model. Not part of StealthMark.

BENCHMARK RESULTS

Compare results

This link opens the same results in the same order.

StealthMark benchmark results

Reset benchmark preferences?

Return to the default view, clear comparison selections and restore the advertisement.