NVIDIAOpen weightsReasoning model

Llama Nemotron Ultra 253B

Not yet ranked: too few independent results so far. Released 7 Apr 2025.

Blended price: –
Output speed: –
First answer: –
Context: 128k

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

Agents & tools36
Knowledge43
Instruction following41
Composite39

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. “Default” means the publisher did not say which setting was used.

Effort	Index	Rank	Speed	First answer	Chat reply cost	Categories
thinkingbest	–	–	–	–	–	Agents & tools 36 · Composite 39 · Instruction following 41 · Knowledge 43
default	–	–	–	–	–	Agents & tools 24 · Coding 52 · Human preference 52 · Reasoning 54

Benchmark results

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.

Reasoning

Benchmark	thinking	default	Source
LMArena Hard Prompts	–	1375#170	LMArena
GPQA Diamond (AA)not in index	72.8%#263	–	Artificial Analysis
Humanity's Last Exam (AA)not in index	7.4%#296	–	Artificial Analysis

Coding

Benchmark	thinking	default	Source
LMArena Coding	–	1391#179	LMArena

Agents & tools

Benchmark	thinking	default	Source
Terminal-Bench Hard	2.3%#304	–	Artificial Analysis
τ²-Bench Telecom (AA)not in index	11.4%#361	–	Artificial Analysis
BFCL Overall	–	10.0%#82	Berkeley Function Calling Leaderboard

Knowledge

Benchmark	thinking	default	Source
AA-Omniscience	-44.9#284	–	Artificial Analysis

Instruction following

Benchmark	thinking	default	Source
IFBench	38.2%#268	–	Artificial Analysis

Human preference

Benchmark	thinking	default	Source
LMArena Text	–	1348#182	LMArena

Composite

Benchmark	thinking	default	Source
AA Intelligence Index	7.5#389	–	Artificial Analysis