reasoning
Llama 3.1 Nemotron Ultra 253B v1
NVIDIA 253B reasoning model derived from Llama 3.1 405B Instruct, ready for commercial use under the NVIDIA Open Model License plus the Llama 3.1 Community License.
Levelmodel
Classreasoning
Parameters253B
Context131,072
Released7 Apr 2025
Opennessopen weights restricted
Runs onnot in the notes
Statusactive
Benchmark Results
vendor-reported
AIME 2025 2025math
Reasoning on, pass@1vendor-reported
72.5
050100
Score in %. Higher is better.
Nemotron Ultra model card
GPQAknowledge and reasoning
Reasoning on, pass@1vendor-reported
76.01
050100
Score in %. Higher is better.
Nemotron Ultra model card
IFEvalknowledge and reasoning
Reasoning on, strict instructionvendor-reported
89.45
050100
Score in %. Higher is better.
Nemotron Ultra model card
LiveCodeBenchcoding
Reasoning on, 2024-08-01 to 2025-02-01 windowvendor-reported
66.31
050100
Score in %. Higher is better.
Nemotron Ultra model card
MATH500math
Reasoning on, pass@1vendor-reported
97
050100
Score in %. Higher is better.
Nemotron Ultra model card
Where this comes from
Every score names who ran it: vendor-reported, maintainer-run or independent. A rank is shown with the population it was ranked in (gap rule 3).
Lineage
distill
Llama-3.1-405B-Instruct is the parent (distill)high confidencemodel cardWhere this comes from
Parent and child relations from model cards, Hugging Face base_model fields and the genealogy tables of the research notes. The tree shows two generations each way, 2 models. Confidence and evidence are the notes' own.
R&D Topics
Identity
Where this comes from
Developer and dates from the research notes. Release date is the vendor's public announcement; weights date is when open weights were published.
Architecture and Size
- Architecture
- dense transformer Derived from Llama-3.1-405B-Instruct by Neural Architecture Search, knowledge distillation (65B tokens) and continual pretraining (88B tokens), then SFT and GRPO
- Total parameters
- 253B sources
- Active parameters
- –
- Checkpoint parameters
- –
- Context length
- 131,072 tokens
- Input
- text
- Output
- text
Where this comes from
Parameter counts as the developer states them. Where the checkpoint holds a different total, both are shown (register A16).
Openness and Access
- Openness
- open weights restricted
- Licence
- NVIDIA Open Model License sources
- Access routes
- open weights download
Where this comes from
The licence file governs over registry tags and aggregator labels.
Deployment
- Runs on
- not in the notes
- Footprint
- not in the notes
- Runtimes
- not in the notes
Where this comes from
Deployment tiers and runtimes as named in the notes. Footprint is the notes' figure, not a measurement by AURION.
Indonesia
- Indonesian support
- not stated
- Regional languages
- not in the notes
- Jakarta hosting
- unknown
- Languages
- English and coding languages; also German, French, Italian, Portuguese, Hindi, Spanish, Thai
Where this comes from
Indonesia fields are carried on every model; unknown is an allowed value.
Industry Use Cases
No industry use case links this model.
Links
DeveloperNVIDIAbig tech, USA
NVIDIAbig tech, USAModelLlama 3.1 Nemotron Ultra 253B v1NVIDIA, reasoning
Llama 3.1 Nemotron Ultra 253B v1NVIDIA, reasoningR&D TopicsReasoning via Large-Scale RL
Use CasesChinese Medical LLMsHealthcareUse cases of models in the same R&D topic; none is recorded for this model itself.
Who Runs Itnone recordedDeployments of these use cases; none names this model.
Organisations 1
Sources
- huggingface.co/nvidia/Llama-3_1-Nemotron-Ultra-253B-v1 https://huggingface.co/nvidia/Llama-3_1-Nemotron-Ultra-253B-v1
To cite this page: Project AURION, AI Use-Cases Research & Innovation, "Llama 3.1 Nemotron Ultra 253B v1", https://aurion.semai.id/models/nvidia-llama-3-1-nemotron-ultra-253b-v1/, records of October 2026.