NVIDIA AI & NIMopen-weight
Released: 2024-10-15

Llama-3.1-Nemotron-70B-Instruct

Custom alignment of Llama 3.1 70B using RLHF and HelpSteer2, topping Arena Hard and AlpacaEval upon release.

Intelligence Index
85.8/ 100
Calibrated multi-domain composite
Throughput Speed
78tok/sec
Output streaming throughput
Time To First Token
330ms
Initial chunk server latency
LiveBench Contam-Free
75.4%
September 2026 suite score

Sourced Benchmark Evaluations (61)

Verified performance across authoritative benchmarks with provenance tracking.
BenchmarkStatusScoreTrustDateSource TypeProvenance
LiveBench (Continuous)Active75.4%
94
2025-02-01independentSource ↗
ARC-AGI-3Active55.9%
93
2024-11-19independentSource ↗
ARC-AGI-2Nearing Saturation67.1%
75
2024-11-12independentSource ↗
FrontierMathActive23.8%
98
2024-11-09independentSource ↗
PutnamBenchActive44.2%
90
2024-11-08independentSource ↗
Omni-MATHDeprecated62.9%
10
2024-11-06independentSource ↗
SWE-LancerDeprecated53.4%
20
2024-11-06independentSource ↗
SoohakActive64.6%
98
2024-11-04independentSource ↗
FrontierCodeActive45.5%
98
2024-11-04independentSource ↗
Humanity's Last ExamActive37.8%
93
2024-11-02independentSource ↗
StructFlowBenchNearing Saturation67.1%
75
2024-11-02independentSource ↗
AIMENearing Saturation80.8%
75
2024-11-02independentSource ↗
OlympiadBenchActive69.7%
80
2024-11-01independentSource ↗
FireBenchActive66.3%
95
2024-10-31independentSource ↗
Terminal-BenchActive55%
80
2024-10-31independentSource ↗
MathArenaActive68%
99
2024-10-31independentSource ↗
OpenAI-MRCRActive72.1%
80
2024-10-30independentSource ↗
CodeContestsDeprecated51%
10
2024-10-30independentSource ↗
SimpleQAActive44.6%
90
2024-10-30independentSource ↗
ComplexBenchNearing Saturation64.7%
80
2024-10-29independentSource ↗
SWE-bench ProActive48.7%
90
2024-10-29independentSource ↗
LongBench-ProActive70.4%
80
2024-10-29independentSource ↗
MultiPL-EDeprecated72.2%
10
2024-10-28independentSource ↗
ZeroSCROLLSNearing Saturation68.6%
78
2024-10-28independentSource ↗
AgentIFNearing Saturation53.7%
80
2024-10-27independentSource ↗
MATHSaturated83.3%
25
2024-10-27independentSource ↗
LiveCodeBenchSaturated64.4%
39
2024-10-27independentSource ↗
InfiniteBenchNearing Saturation72.9%
65
2024-10-27independentSource ↗
MMLU-ProNearing Saturation78.9%
75
2024-10-26independentSource ↗
MercuryDeprecated69.1%
10
2024-10-26independentSource ↗
SCROLLSNearing Saturation68.6%
65
2024-10-26independentSource ↗
RULERNearing Saturation78.9%
84
2024-10-25independentSource ↗
TabMWPSaturated78.2%
25
2024-10-25independentSource ↗
CyberSecEvalDeprecated70.7%
10
2024-10-25independentSource ↗
MMLU-ReduxSaturated84.1%
25
2024-10-24independentSource ↗
QuALITYNearing Saturation75.5%
78
2024-10-24independentSource ↗
SWE-bench VerifiedSaturated61.2%
25
2024-10-24independentSource ↗
MGSMNearing Saturation80.8%
65
2024-10-23independentSource ↗
LongBenchNearing Saturation73.8%
65
2024-10-23independentSource ↗
SWE-benchDeprecated53.4%
10
2024-10-23independentSource ↗
AGIEvalSaturated74%
25
2024-10-22independentSource ↗
QASPERNearing Saturation71.2%
65
2024-10-22independentSource ↗
TruthfulQASaturated75.5%
25
2024-10-22independentSource ↗
SuperGLUESaturated87.7%
25
2024-10-21independentSource ↗
GSM-HardDeprecated81.6%
10
2024-10-21independentSource ↗
MT-BenchNearing Saturation8.1%
75
2024-10-20independentSource ↗
BIG-Bench HardSaturated84.3%
25
2024-10-20independentSource ↗
CodeXGLUEDeprecated73.8%
23
2024-10-20independentSource ↗
MBPP+Nearing Saturation80.1%
65
2024-10-19independentSource ↗
WinoGrandeSaturated89.4%
28
2024-10-19independentSource ↗
IFEvalDeprecated77.4%
25
2024-10-19independentSource ↗
GPQANearing Saturation75.5%
78
2024-10-19independentSource ↗
HumanEval+Saturated84.8%
25
2024-10-18independentSource ↗
Needle-in-a-HaystackSaturated97.2%
44
2024-10-18independentSource ↗
CommonsenseQASaturated90.3%
28
2024-10-18independentSource ↗
GSM8KSaturated91.8%
25
2024-10-17independentSource ↗
GLUESaturated92.9%
28
2024-10-17independentSource ↗
MBPPSaturated83.2%
25
2024-10-17independentSource ↗
HumanEvalSaturated89.5%
25
2024-10-16independentSource ↗
MMLUSaturated87.5%
25
2024-10-16independentSource ↗
ARC-AGISaturated80.8%
25
2024-10-16independentSource ↗
Local Hardware Execution (Ollama)Hardware Compatible
# 1. Pull and execute model locally with Ollama
ollama run llama-3-1-nemotron-70b-instruct

# 2. Or invoke via local OpenAI-compatible endpoint
curl http://localhost:11434/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "llama-3-1-nemotron-70b-instruct",
    "messages": [{"role": "user", "content": "Analyze reasoning chains on AIME 2026"}]
  }'

Evaluation & Sourcing Notes

Scores listed for Llama-3.1-Nemotron-70B-Instruct represent verified evaluations extracted from official research papers, independent evaluation suites (HELM, LMSYS, OpenCompass, LiveBench), and verified audit reports.

All benchmarks marked as saturated or deprecated reflect historical performance where the benchmark no longer provides active discriminative power.