Llama 3.1 70B Instruct vs Llama 3.1 Nemotron Instruct 70B
Wins 1 of 6 areas
Following instructions
Wins 1 of 6 areas
Reasoning
The two are evenly matched.Llama 3.1 70B Instruct is better at following instructions; Llama 3.1 Nemotron Instruct 70B at reasoning.
Scores updated · 21 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.
- Following instructionsDoing exactly what it is asked10Llama 3.1 70B Instruct1 of 1 test
- ReasoningHard problems that need careful thinking01Llama 3.1 Nemotron Instruct 70B1 of 3 tests · 2 ties
- CodingWriting and fixing software11Even1 each
- AgentsCarrying out multi-step tasks on its own00Even0 each · 1 tie
- FactsGetting facts right instead of making them up11Even1 each
- Long documentsFinding answers in very long texts00Even0 each · 1 tie
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 three tests each model wins by the widest margin. Scores are out of 100.
Where Llama 3.1 70B Instruct pulls ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+3.7points ahead
- Code for real scientific research problemsSciCode+3.4points ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+1.9points ahead
Where Llama 3.1 Nemotron Instruct 70B pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+6.9points ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+5.6points ahead
- Hard command-line tasks in a real terminalTerminal-Bench Hard+1.5points ahead
Every test, side by side
All 21 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingEven
- SciCodeLlama 3.1 70B Instruct by 3.426.723.3+3.4
- Terminal-Bench HardLlama 3.1 Nemotron Instruct 70B by 1.534.5+1.5
ReasoningLlama 3.1 Nemotron Instruct 70B
- GPQA DiamondLlama 3.1 Nemotron Instruct 70B by 5.640.946.5+5.6
- Humanity's Last Examtie4.54.2tie
- CritPttie00tie
FactsEven
- AA-Omniscience · Non-hallucinationLlama 3.1 Nemotron Instruct 70B by 6.921.828.7+6.9
- AA-Omniscience · AccuracyLlama 3.1 70B Instruct by 1.919.717.8+1.9
Following instructionsLlama 3.1 70B Instruct
- IFBenchLlama 3.1 70B Instruct by 3.734.430.7+3.7
Other results11 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.
- Arena HardLlama 3.1 Nemotron Instruct 70B by 29.355.785+29.3
- ARC-ChallengeLlama 3.1 70B Instruct by 25.694.869.2+25.6
- AlpacaEval 2 LCLlama 3.1 Nemotron Instruct 70B by 19.538.157.6+19.5
- MT-BenchLlama 3.1 70B Instruct by 8.18.20.1+8.1
- τ²-Bench Telecom (AA run)Llama 3.1 Nemotron Instruct 70B by 7.915.223.1+7.9
- AA Agentic IndexLlama 3.1 Nemotron Instruct 70B by 2.65.17.7+2.6
- GSM8KLlama 3.1 70B Instruct by 2.493.891.4+2.4
- AA-OmniscienceLlama 3.1 Nemotron Instruct 70B by 2.3-43.1-40.8+2.3
- WinoGrandetie85.384.5tie
- AA Intelligencetie6.66.9tie
- Artificial Analysis Coding Indextie10.910.8tie
Questions people ask
Which is better, Llama 3.1 70B Instruct or Llama 3.1 Nemotron Instruct 70B?
Llama 3.1 70B Instruct and Llama 3.1 Nemotron Instruct 70B each win one of the six areas where both have results. Llama 3.1 70B Instruct wins following instructions; Llama 3.1 Nemotron Instruct 70B wins reasoning. They are level on coding, agents, facts and long documents.
Which is better for coding?
Neither. They win 1 coding test each of the 2 both models report.
How do you compare the two?
We use the 21 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 11 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.