What are Nemotron 3 Ultra 550B A55B's benchmark scores?

As of 25 Oct 2026, Nemotron 3 Ultra 550B A55B (thinking effort) scores 55.7 ±11.3 on the BenchLeader Index, strongest in instruction following (79) and weakest in agents & tools (38). That rests on 10 independent benchmark results, and every score on this page links to its publisher.

How does Nemotron 3 Ultra 550B A55B rank?

Nemotron 3 Ultra 550B A55B (thinking effort) is #222 of 785 ranked configurations on 25 Oct 2026. 509 other configurations score within its margin, so the exact position matters less than the band. The leader is Claude Opus 5.5 (max effort) at 72.6.

How much does Nemotron 3 Ultra 550B A55B cost per 1M tokens?

As of 25 Oct 2026, Nemotron 3 Ultra 550B A55B costs $0.600 per 1M input tokens and $2.70 per 1M output tokens, $1.13 blended (three input tokens per output token). The list price last changed on 17 Oct 2026.

When was Nemotron 3 Ultra 550B A55B released?

NVIDIA released Nemotron 3 Ultra 550B A55B on 4 Jun 2026. BenchLeader has results for it at 2 reasoning-effort settings.

NVIDIA Open weights: nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16Reasoning modelCompare with another model

Nemotron 3 Ultra 550B A55B

Reasoning effort

Nemotron 3 Ultra 550B A55B is a NVIDIA open-weights reasoning model, released 4 Jun 2026. Its best configuration (thinking reasoning effort) ranks #222 of 785 on the BenchLeader Index at 55.7 ±11.3, in the upper half. The ± is the point: 509 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in instruction following (79) and lowest in agents & tools (38). At $1.13 per million tokens blended it is pricier than most ranked models. Output speed of 154 tokens per second puts it in the fastest quarter, with a first answer in 16.0 s. It has been measured at 2 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 4 Jun 2026.

Blended price: $1.13/M; $0.600 in · $2.70 out
Output speed: 154 tok/s; measured by Artificial Analysis
First answer: 16 s; first token 0.34 s
Context: 1M
Full answer: 19 s; median, reasoning included
Cost per run: $0.470; one full Intelligence Index run
Released: 4 Jun 2026

How it scores by categoryDashed line = average model (50). One step of 15 = one standard deviation.

Overall index56
Reasoning46
Agents & tools38
Knowledge61
Instruction following79
Long context65
Composite55

Reasoning-effort configurations

The same model behaves differently depending on how much it is allowed to think. Each row is one setting, scored only on the benchmarks that were run at that setting. “Not stated” collects results from publishers that did not say which setting they used; for a reasoning model that is usually its thinking mode, but we do not assume it. Pick a setting here or at the top of the page to see its price, speed and category scores.

Effort	Index	Rank	Speed	First answer	Chat reply cost	Categories
thinkingbest	55.7	#222	154 tok/s	16 s	$0.0011	Agents & tools 38 · Composite 55 · Instruction following 79 · Knowledge 61 · Long context 65 · Reasoning 46
not stated	55.5	#229	91 tok/s	0.34 s	$0.0011	Agents & tools 40 · Coding 62 · Human preference 61 · Knowledge 58 · Maths 62 · Reasoning 60

How its index has moved

9 Sept 2026 to 25 Oct 2026

thinking
not stated

The index is recomputed from scratch every day, so a line moves when a new benchmark result lands, when a publisher revises a score, or when the models it is normalised against change. Early movement usually means the score is still settling.

Benchmark results

See all LLM benchmarks. One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.

Metadata

Benchmark	thinking	Source	Trend
AA Openness Indexnot in index	83.3#12	Artificial Analysis

Reasoning

Benchmark	thinking	not stated	Source
Arena Hard Promptseffort not stated	–	1444#119	Arena (LMArena)
LiveBench 1.0 Reasoningnot in indexeffort not stated	–	74.7%#61	LiveBench
GPQA Diamond (AA)not in index	86.7%#107	–	Artificial Analysis
Humanity's Last Exam (AA)not in index	28.4%#164	–	Artificial Analysis
GPQA Diamond (Vals)not in indexeffort not stated	–	86.1%#42	Vals AI
CritPt	3.1%#176	–	Artificial Analysis

Coding

Benchmark	thinking	not stated	Source
Arena Codingeffort not stated	–	1473#113	Arena (LMArena)
LiveBench 1.0 Codingnot in indexeffort not stated	–	70.7%#61	LiveBench
SciCode (AA)not in index	40.3%#171	–	Artificial Analysis
LiveCodeBencheffort not stated	–	86.0%#23	Vals AI
SWE-bench (Vals)not in indexeffort not stated	–	69.0%#64	Vals AI
Code Migrationnot in indexeffort not stated	–	4.9%#69	Vals AI
ProgramBenchnot in indexeffort not stated	–	0.0%#32	Vals AI
Vibe Code Bench v1.1not in indexeffort not stated	–	7.6%#99	Vals AI
Terminal-Bench 4.0 (AA)not in index	0.5%#139	–	Artificial Analysis
Terminal-Bench 2.1 (AA)not in index	53.9%#110	–	Artificial Analysis
FrontierCode 1.1 (AA)not in index	10.5%#65	–	Artificial Analysis

Agents & tools

Benchmark	thinking	not stated	Source
LiveBench Agentic Codingnot in indexeffort not stated	–	38.7%#65	LiveBench
Terminal-Bench Hard	36.4%#71	–	Artificial Analysis
τ²-Bench Telecom (AA)not in index	83.3%#110	–	Artificial Analysis
Terminal-Bench 2.1 (Vals)effort not stated	–	50.9%#56	Vals AI
GDPval-AA v2.1not in index	25.6%#94	–	Artificial Analysis
τ³-Banking (AA)	14.2%#123	–	Artificial Analysis
Analyst Agent (AA)not in index	6.3%#49	–	Artificial Analysis
Finance Agent v2not in indexeffort not stated	–	37.7%#65	Vals AI
Harvey's Legal Agent Benchmark v1not in indexeffort not stated	–	0.4%#58	Vals AI
Legal Research Benchnot in indexeffort not stated	–	15.4%#65	Vals AI
AutomationBenchnot in index	3.0%#176	–	Artificial Analysis
GDP.pdf (AA)not in index	5.0%#166	–	Artificial Analysis
EnterpriseOps-Gymnot in index	28.9%#43	–	Artificial Analysis
AA-Briefcase v1.1not in index	876#51	–	Artificial Analysis
Terminal-Bench-Science 0.1 (AA)not in index	0.0%#72	–	Artificial Analysis
CWE-Bench (AA)	27.5%#20	–	Artificial Analysis
DeepSec-Bench (AA)	1.1%#22	–	Artificial Analysis
CyberGym end-to-end (AA)	10.7%#14	–	Artificial Analysis

Maths

Benchmark	not stated	Source	Trend
LiveBench Mathematicsnot in index	88.7%#42	LiveBench
Arena Maths	1444#85	Arena (LMArena)

Knowledge

Benchmark	thinking	not stated	Source
LiveBench 1.0 Data Analysisnot in indexeffort not stated	–	54.5%#63	LiveBench
AA-Omniscience	-0.4#139	–	Artificial Analysis
MMLU-Proeffort not stated	–	85.8%#53	Vals AI
LegalBencheffort not stated	–	82.1%#74	Vals AI
CorpFineffort not stated	–	65.5%#27	Vals AI
TaxEvaleffort not stated	–	73.1%#51	Vals AI
Excel Modeling Benchmarknot in indexeffort not stated	–	31.7%#62	Vals AI
MedCodenot in indexeffort not stated	–	38.6%#68	Vals AI
AA-Omniscience: accuracynot in index	22.6%#272	–	Artificial Analysis
AA-Omniscience: non-hallucinationnot in index	70.3%#41	–	Artificial Analysis

Instruction following

Benchmark	thinking	not stated	Source
LiveBench Languagenot in indexeffort not stated	–	70.8%#60	LiveBench
LiveBench Instruction Followingnot in indexeffort not stated	–	73.5%#16	LiveBench
IFBench	81.4%#3	–	Artificial Analysis
Arena Instruction Followingnot in indexeffort not stated	–	1405#130	Arena (LMArena)

Human preference

Benchmark	not stated	Source
Arena Text	1426#116	Arena (LMArena)
Arena Creative Writingnot in index	1382#133	Arena (LMArena)
Arena Multi-turnnot in index	1398#155	Arena (LMArena)
Arena Longer Queriesnot in index	1433#118	Arena (LMArena)

Multimodal

Benchmark	thinking	Source	Trend
AA-3Dnot in index	573#22	Artificial Analysis

Long context

Benchmark	thinking	Source	Trend
AA-LCR	79.3%#98	Artificial Analysis
MLCRnot in index	5.6%#140	Artificial Analysis

Composite

Benchmark	thinking	not stated	Source
LiveBench 1.0not in indexeffort not stated	–	67.4%#61	LiveBench
AA Intelligence Index v4.3.2	22.9#192	–	Artificial Analysis
AA Cyber Indexnot in index	13.1#19	–	Artificial Analysis

Where it wins

Benchmarks where this configuration ranks in the top five of every configuration measured.

#3IFBench81.4% · of 423

Where to run it

Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.

Provider	Speed	First token	Input $/M	Output $/M	Context	Quantisation
Baseten	94 tok/s	0.29 s	$0.600	$2.40	203k	fp4
Baseten	91 tok/s	0.33 s	$0.600	$2.40	203k	fp4
DeepInfra	79 tok/s	1.15 s	$0.500	$2.20	262k	fp4
Venice	64 tok/s	1.29 s	$0.625	$3.13	256k	fp8

Price history

The price above, per 1M tokens, every day it changed.

input outputnow $0.600 in · $2.70 out

What a task costs

Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache at $0.120 per 1M. Reasoning tokens are not modelled.

Workload	Tokens in / out	Cost	With caching	Time
Chat reply	400 / 300	$0.0011	$0.0009	17.9 s
Summarize a 30-page report	12,000 / 600	$0.0088	$0.0045	19.9 s
Code edit	6,000 / 1,500	$0.0077	$0.0055	25.7 s
Agentic coding session	60,000 / 4,000	$0.047	$0.025	41.9 s
Structured extraction	2,000 / 200	$0.0017	$0.0010	17.3 s

New this month

Questions people ask

What are Nemotron 3 Ultra 550B A55B's benchmark scores?: As of 25 Oct 2026, Nemotron 3 Ultra 550B A55B (thinking effort) scores 55.7 ±11.3 on the BenchLeader Index, strongest in instruction following (79) and weakest in agents & tools (38). That rests on 10 independent benchmark results, and every score on this page links to its publisher.
How does Nemotron 3 Ultra 550B A55B rank?: Nemotron 3 Ultra 550B A55B (thinking effort) is #222 of 785 ranked configurations on 25 Oct 2026. 509 other configurations score within its margin, so the exact position matters less than the band. The leader is Claude Opus 5.5 (max effort) at 72.6.
How much does Nemotron 3 Ultra 550B A55B cost per 1M tokens?: As of 25 Oct 2026, Nemotron 3 Ultra 550B A55B costs $0.600 per 1M input tokens and $2.70 per 1M output tokens, $1.13 blended (three input tokens per output token). The list price last changed on 17 Oct 2026.
When was Nemotron 3 Ultra 550B A55B released?: NVIDIA released Nemotron 3 Ultra 550B A55B on 4 Jun 2026. BenchLeader has results for it at 2 reasoning-effort settings.

Data as of 25 Oct 2026. Compare these configurations.

Cite as: BenchLeader, “Nemotron 3 Ultra 550B A55B: benchmarks, pricing, speed and rank”, data as of 25 Oct 2026, https://www.benchleader.com/models/nemotron-3-ultra-550b-a55b.