Llama 3.1 Nemotron Ultra 253B V1 vs NVIDIA Nemotron Nano 9B V2
Wins 3 of 6 areas
Coding · Reasoning · Following instructions
Wins 1 of 6 areas
Long documents
Llama 3.1 Nemotron Ultra 253B V1 wins more areas, narrowly.NVIDIA Nemotron Nano 9B V2 is better at long documents.
Scores updated · 19 tests both models report · How we compare
Where each one wins
Tests won in each of the six areas where both have results. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.
- ReasoningHard problems that need careful thinking20Llama 3.1 Nemotron Ultra 253B V12 of 3 tests · 1 tie
- CodingWriting and fixing software10Llama 3.1 Nemotron Ultra 253B V11 of 2 tests · 1 tie
- Following instructionsDoing exactly what it is asked10Llama 3.1 Nemotron Ultra 253B V11 of 1 test
- Long documentsFinding answers in very long texts01NVIDIA Nemotron Nano 9B V21 of 1 test
- AgentsCarrying out multi-step tasks on its own00Even0 each · 1 tie
- FactsGetting facts right instead of making them up11Even1 each
Agents, long documents and following instructions rest on a single test each.
What it costs
Prices per million tokens, roughly 750,000 words. The bars show the cost of a million tokens read plus a million written.
The biggest differences
The tests each model wins by the widest margin, up to three each. Scores are out of 100.
Where Llama 3.1 Nemotron Ultra 253B V1 pulls ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+15.8points ahead
- Code for real scientific research problemsSciCode+12.7points ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+10.6points ahead
Where NVIDIA Nemotron Nano 9B V2 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+20.7points ahead
- Reasons across sets of long documentsAA-LCR+15points ahead
Every test, side by side
All 19 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingLlama 3.1 Nemotron Ultra 253B V1
- SciCodeLlama 3.1 Nemotron Ultra 253B V1 by 12.734.722+12.7
- Terminal-Bench Hardtie2.31.5tie
ReasoningLlama 3.1 Nemotron Ultra 253B V1
- GPQA DiamondLlama 3.1 Nemotron Ultra 253B V1 by 15.872.857+15.8
- Humanity's Last ExamLlama 3.1 Nemotron Ultra 253B V1 by 2.57.44.9+2.5
- CritPttie00tie
FactsEven
- AA-Omniscience · Non-hallucinationNVIDIA Nemotron Nano 9B V2 by 20.718.839.5+20.7
- AA-Omniscience · AccuracyLlama 3.1 Nemotron Ultra 253B V1 by 8.320.111.8+8.3
Long documentsNVIDIA Nemotron Nano 9B V2
- AA-LCRNVIDIA Nemotron Nano 9B V2 by 157.722.7+15
Following instructionsLlama 3.1 Nemotron Ultra 253B V1
- IFBenchLlama 3.1 Nemotron Ultra 253B V1 by 10.638.227.6+10.6
Other results9 tests, not counted
Tests outside the eight areas. They are not counted above: several are summary scores built from other tests, or the same test under another name.
- τ²-Bench Telecom (AA run)NVIDIA Nemotron Nano 9B V2 by 10.511.421.9+10.5
- AA Agentic IndexNVIDIA Nemotron Nano 9B V2 by 5.63.89.4+5.6
- Artificial Analysis Coding IndexLlama 3.1 Nemotron Ultra 253B V1 by 4.813.18.3+4.8
- LiveCodeBenchNVIDIA Nemotron Nano 9B V2 by 4.866.371.1+4.8
- AA-OmniscienceNVIDIA Nemotron Nano 9B V2 by 3.4-44.9-41.5+3.4
- IFEvaltie89.590.3tie
- MATH-500 (EM)tie9797.8tie
- AIME 2025tie72.572.1tie
- AA Intelligencetie7.57.4tie
Questions people ask
Which is better, Llama 3.1 Nemotron Ultra 253B V1 or NVIDIA Nemotron Nano 9B V2?
Llama 3.1 Nemotron Ultra 253B V1 wins three of the six areas where both have results: coding, reasoning and following instructions. NVIDIA Nemotron Nano 9B V2 wins long documents. They are level on agents and facts.
Which is better for coding?
Llama 3.1 Nemotron Ultra 253B V1. It wins 1 of the 2 coding tests both models report; NVIDIA Nemotron Nano 9B V2 wins none, and 1 is a tie.
How do you compare the two?
We use the 19 benchmark tests both models have published scores on. The verdict counts the 10 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 9 are listed but not counted, because several are summary scores or repeat a test. Each score is the one shown on the model's own page.