Nemotron 3 Super nemotron-3-super-120b-a12b

nvidia · nvidia

unranked — no quality signals yet

params
—
arch
moe
context
256k
license
— open weights

# architecture

attn shared GQA 32q/2kv router top-22 of 512 +1 shared expert ×512 expert out ×88 layers ctx 262,144
layers
88
d_model
4096
heads
32q / 2kv
head dim
128
experts
top-22 of 512 + 1 shared
vocab
131072
family
nemotron_h

signal path of one layer, generated from the registry's structured fields — dimension lines quote the model's real numbers; an MoE trace forks at the router, a dense trace runs straight through. Geometry fields come from the repo's config.json.

# why ranked

Overall: provisional — score 50.7 ● low — a single signal; treat with caution

provisional: too few independent signals for a numbered position — the score is shown, but this model sorts after every ranked model.

signalweightinput (0–100)
aa_intelligence 0.70 —
bench_composite 0.30 50.7

benchmark panel evidence:

complete panelscore (0–100)
aime-2025-v169.0
browsecomp-v132.5

missing signals are dropped and the remaining weights renormalized — never imputed.

full methodology

# trend

OGM score · last 45 days
OGM score over 45 days

# benchmarks

benchmarkscoresourcedate
Aa-lcr 0.6 / 1 LLM Stats —
AIME 2025 0.9 / 1 LLM Stats —
Arena-hard-v2 0.7 / 1 LLM Stats —
Bird-sql-(dev) 0.4 / 1 LLM Stats —
Browsecomp 0.3 / 1 LLM Stats —

# where to run

providerquantctx$/M in$/M out$/M cacheprice srctpsuptime✓
Nebius — ——— nebius — — ✓
Nvidia 1024k ——— via requesty — —
DeepInfrabf16 256k $0.09$0.40— via openrouter — 100.0%
DekaLLMfp8 256k $0.08$0.45— via openrouter — 99.7%
Crusoefp8 256k $0.30$2.40$0.15 via openrouter — 100.0%

sorted by blended price ((3·input + output) / 4 per 1M) · ✓ = the provider's own catalog confirms the offer · "via …" prices are what the aggregator routing the offer charges, not the provider's own list price

source aliases
llmstats
nemotron-3-super-120b-a12b
openrouter
nvidia/nemotron-3-super-120b-a12b