models · nvidia

NVIDIA Nemotron 3 Nano 30B A3B

NVIDIA Nemotron 3 Nano 30B A3B — catalogued from third-party evaluation data; no weights held by Sovereign Frontier.

What this registry has checked

Closed. A third party reports this model and we have recorded their numbers. Nothing of our own: no weights seen, no publisher hash held. Treat every figure here as their claim.
Methodthird-party-report

What is recorded

Publishernvidia
OriginUnited States
LicenceNVIDIA-Open-Model
Pulls recorded here6

Reported scores

MMLU-Pro57.9
GPQA Diamond39.9
AIME 202513.3
LiveCodeBench36
IFBench37.5
Humanity's Last Exam4.6

Reported by a third party and recorded here, not re-run by us. Source: artificialanalysis.ai/api/v2.

Measured on real hardware

None of these runs are ours. Each figure is the reporter’s, linked to their own evidence, under their own trust label. The number shown is single-stream decode — one user, concurrency 1 — because aggregate throughput across many users is a different quantity and reads as far faster than anyone will see.

Measured by local.ai

HardwareBuild measuredDecode @ 8KMemory @ 8KDecode mode
NVIDIA GeForce RTX 5090 32GBNVFP4417 tok/s30.4 GBordinary
NVIDIA RTX PRO 6000 Blackwell 96GBNVFP4317 tok/s90.6 GBordinary
NVIDIA RTX PRO 6000 Blackwell 96GBFP8254 tok/s90.3 GBordinary
NVIDIA RTX 6000 Ada 48GBNVFP4219 tok/s45.6 GBordinary
NVIDIA RTX 6000 Ada 48GBFP8187 tok/s45.4 GBordinary
NVIDIA DGX Spark 128GBNVFP460.5 tok/s54.7 GBordinary
NVIDIA DGX Spark 128GBFP844.1 tok/s112.6 GBordinary
MacBook Pro M5 Max 128GB · 40-core GPUNVFP4109 tok/s22.9 GBnot stated

Measured by local.ai, not by us: 13 runs, 5 more not listed. Each names its engine pinned by image digest and the exact serve command, in the full record.

Elsewhere in the registry

Registry record as of 2026-07-26. Identity, licence and parameter count are properties of a release and do not move; tier can, and the verify link above re-checks it against the live registry rather than this page.