NVIDIA: Nemotron 3 Ultra
NVIDIA: Nemotron 3 Ultra is nvidia's model. On the un-leaked meo benchmark it ranks #13 of 22 with 50.3% accuracy and an Effective Value (๐) of 0.0004 โ earning a correct answer for about $0.0496 in 147s each.
as of 2026-06-08
meo rank
#13
of 22
Accuracy
50.3%
overall
๐ index (top=100)
0.73
raw ๐ 0.0004 ยท N=10
$ / correct
$0.0496
147s each
Per-domain accuracy
Tape machine
100%
Theory of mind
100%
CSP (unsatisfiable)
94%
Regex automaton
94%
Framework bias
67%
Long arithmetic
7%
Logic / math / CS
5%
Base-rate bias
2%
Watson-Glaser
0%
State tracking
0%
Efficiency
- Cost per correct
- $0.0496
- Seconds per correct
- 147
- Tokens per correct
- 19,710
- Correct answers
- 54 / 151
- Total benchmark cost
- $2.68
- Avg tokens / item
- 7,049
Model details
- Lab
- nvidia
- Released
- 2026-06-04
- Weights
- Open
- Context
- 1M
- Max output
- 16K
- Knowledge cutoff
- โ
- Price in
- $0.50/M
- Price out
- $2.50/M
- Modality
- text->text
Sources: openrouter ยท meo
FAQ
How accurate is NVIDIA: Nemotron 3 Ultra?
NVIDIA: Nemotron 3 Ultra scores 50.3% overall on the meo benchmark (rank #13 of 22), an un-leaked private holdout graded objective-first.
What is NVIDIA: Nemotron 3 Ultra's Effective Value?
Its Effective Value (๐) is 0.0004 at N=10, ฯ=1, ฮด=1.5 โ a fused metric of accuracy, speed, cost, and error-cascade. ๐ rankings shift with task depth.
How much does NVIDIA: Nemotron 3 Ultra cost per correct answer?
About $0.0496 per correct answer (19,710 tokens, 147 seconds each) at benchmark-time pricing via OpenRouter.
Is NVIDIA: Nemotron 3 Ultra open weights?
Yes โ NVIDIA: Nemotron 3 Ultra is released with open weights.