NVIDIAUnited States· released Jun 2026· open weights

Nemotron 3 Ultra

Name: Nemotron 3 Ultra
Author: NVIDIA

Provisional: 8 comparable tests so far. Ranked once it reaches 6 across 3 counting categories.

NVIDIA, United States. One of 2 models from this lab on the Index.

Third-party publishedLow confidence (thin coverage)

Snapshot September 8, 2026

OpenCharts Index

49.8

Provisional: not enough comparable tests or counting categories to rank yet · 8 comparable tests · 0 counting categories

Drop any single test and the Index lands between 41.3 and 58.2.

Needs more results (0 of 8 so far)

Distance to today's frontier on the hardest tests. Not a measure of intelligence.

What it's known for

Nemotron 3 Ultra, in 8 results.

Its widest gap is on ProofBench, 98.0% behind the best published result. No comparable results yet in knowledge, multimodal and long context. Between 50% and 60% behind the best published results, on average.

Strongest category

No category counts yet: a category counts once it has more than one comparable test and the model has taken at least half of them. The scores it has are below.

Best single result

Mock AIME 2024–2025

Solve competition mathematics problems that have a single exact integer answer.

86.7%

86.7% of the best (100.0%) · #49 of 78.

Olympiad-style problems need long, careful chains of reasoning, and there is no partial credit.

Where it trails

ProofBench

Write a complete, rigorous proof, graded for rigor rather than for a final answer.

2.0%

98.0% behind the best published result (100.0%) · #60 of 61.

A proof shows the reasoning itself. A right answer with a wrong argument scores nothing.

Not measured yet

No comparable result yet in these 3 categories. A blank is a blank, never a zero, and it does not lower the Index.

Knowledge
Multimodal
Long context

Covered, but not averaged into the Index:

Reasoning2 of 6 tests · needs 3 · mean 47.5
Coding2 of 7 tests · needs 4 · mean 55.6
Math2 of 5 tests · needs 3 · mean 44.4
Agentic1 of 8 tests · needs 4 · mean 24.3
Human preferencesingle-test category, shown beside the Index · 77.1

Every score, with its source

See it at work

What Nemotron 3 Ultra was asked to do, and how it did.

Every test it has taken, by category, with the kind of task it faced, who produced the number and where the result landed against the best published one. The examples are original and illustrative, never items from the datasets themselves.

Human preference · mean 77.1 across 1 comparable test · a single-test category, shown beside the Index and never averaged into it

Text ArenaLeaderboard

Answer the same prompt as a rival model; real users vote blind on which answer they prefer.

1,426

39% expected win rate against the board leader (1,507)
#70 of 92

board 2026-09-02

Source: Arena

Every result is read against the best published one on its test: percent scores as a share of the best above chance, Arena ratings as the expected win rate against the board leader, open-ended values on a log scale. Sitting a harder exam never lowers a model. Every number links to the publisher that produced it.

Head to head

Nemotron 3 Ultra against whoever you pick.

Choose a rival. Every counted test both have taken appears side by side as a share of the best published result on that test, with the category means above. A win is a higher share; a gap under 2 points is a tie, because one item on a 45-problem exam is 2.2 points and publishers' own standard errors are of that order.

Nemotron 3 Ultra

provisional·Index 49.8D

0 : 7

wins · 0 ties · 7 shared

GPT-6 Astra

#1·Index 99.0A+

GPT-6 Astra comes out ahead on 7 of the 7 tests both have taken.

Category means

Reasoning
47.5
99.6
Knowledge
—
100.0
Coding
55.6
97.7
Math
44.4
99.8
Agentic
24.3
98.5
Human preference
77.1
—

Shared tests · biggest gaps first

ProofBenchMath
2.0%
99.0%
−97.0
CritPtReasoning
3.1%
31.7%
−88.5
APEX-AgentsAgentic
11.5%
46.7%
−74.2
WeirdMLCoding
43.5%
92.9%
−53.2
SciCodeCoding
39.9%
56.5%
−26.6
GPQA DiamondReasoning
85.4%
95.8%
−14.7
Mock AIME 2024–2025Math
86.7%
100.0%
−13.3

Bars are distances to the best published result on each test (percent scores above chance, Arena ratings as win rate against the leader, open-ended values on a log scale). Raw values as the publishers report them.

Open GPT-6 Astra

Scores

Every test, every source

The best published run per benchmark, the raw value as the publisher reports it, who produced the number, how close it comes to the best published result on that test, and where the model sits among every model scored on it. Percent scores are read above chance, Arena ratings as a win rate against the board leader, open-ended values on a log scale. A test with too few models to compare against is shown but not counted.

Category means

Reasoning47.5 · 2/6
Knowledge—
Coding55.6 · 2/7
Math44.4 · 2/5
Agentic24.3 · 1/8
Multimodal—
Human preferencebeside77.1 · 1/1
Long context—

Mean · tests taken of the category's comparable tests. A faded category either rests on fewer than half of them or is a single-test category (marked “beside”), which is shown next to the Index and never averaged into it.

GPQA DiamondEpoch-runReasoning

85.4%

85.3% of the best (95.8%), above chance
32th percentile of 78 · n = 198

Source: Epoch AI — AI Benchmarking Hub

CritPtLeaderboardReasoning

3.1%

9.7% of the best (32.3%)
32th percentile of 78

Source: Epoch AI — AI Benchmarking Hub

SciCodeLeaderboardCoding

39.9%

64.4% of the best (62.0%)
18th percentile of 73 · n = 338

Source: Epoch AI — AI Benchmarking Hub