NVIDIA AI & NIMopen-weight
Released: 2024-06-14

Nemotron-4-340B-Instruct

Massive 340B parameter dense model developed for synthetic data generation and LLM pipeline training across industry verticals.

Intelligence Index
84.5/ 100
Calibrated multi-domain composite
Throughput Speed
42tok/sec
Output streaming throughput
Time To First Token
510ms
Initial chunk server latency
LiveBench Contam-Free
73.0%
September 2026 suite score

Sourced Benchmark Evaluations (51)

Verified performance across authoritative benchmarks with provenance tracking.
BenchmarkStatusScoreTrustDateSource TypeProvenance
LiveBench (Continuous)Active73%
94
2025-02-01independentSource ↗
ARC-AGI-3Active54.6%
93
2024-07-19independentSource ↗
ARC-AGI-2Nearing Saturation65.5%
75
2024-07-12independentSource ↗
FrontierMathActive23%
98
2024-07-09independentSource ↗
PutnamBenchActive42.6%
90
2024-07-08independentSource ↗
SWE-LancerDeprecated51.7%
20
2024-07-06independentSource ↗
Omni-MATHDeprecated60.7%
10
2024-07-06independentSource ↗
SoohakActive62.3%
98
2024-07-04independentSource ↗
FrontierCodeActive44.1%
98
2024-07-04independentSource ↗
AIMENearing Saturation77.9%
75
2024-07-02independentSource ↗
StructFlowBenchNearing Saturation66.1%
75
2024-07-02independentSource ↗
Humanity's Last ExamActive37%
93
2024-07-02independentSource ↗
OlympiadBenchActive67.2%
80
2024-07-01independentSource ↗
MathArenaActive65.6%
99
2024-06-30independentSource ↗
Terminal-BenchActive53.2%
80
2024-06-30independentSource ↗
FireBenchActive65.3%
95
2024-06-30independentSource ↗
SimpleQAActive43.9%
90
2024-06-29independentSource ↗
CodeContestsDeprecated49.4%
10
2024-06-29independentSource ↗
ComplexBenchNearing Saturation63.7%
80
2024-06-28independentSource ↗
SWE-bench ProActive47.1%
90
2024-06-28independentSource ↗
MultiPL-EDeprecated69.9%
10
2024-06-27independentSource ↗
LiveCodeBenchSaturated62.3%
39
2024-06-26independentSource ↗
MATHSaturated80.4%
25
2024-06-26independentSource ↗
AgentIFNearing Saturation52.9%
80
2024-06-26independentSource ↗
MercuryDeprecated66.9%
10
2024-06-25independentSource ↗
MMLU-ProNearing Saturation77.7%
75
2024-06-25independentSource ↗
CyberSecEvalDeprecated68.4%
10
2024-06-24independentSource ↗
TabMWPSaturated75.4%
25
2024-06-24independentSource ↗
SWE-bench VerifiedSaturated59.3%
25
2024-06-23independentSource ↗
MMLU-ReduxSaturated82.8%
25
2024-06-23independentSource ↗
SWE-benchDeprecated51.7%
10
2024-06-22independentSource ↗
MGSMNearing Saturation77.9%
65
2024-06-22independentSource ↗
TruthfulQASaturated74.4%
25
2024-06-21independentSource ↗
AGIEvalSaturated72.2%
25
2024-06-21independentSource ↗
SuperGLUESaturated85.7%
25
2024-06-20independentSource ↗
GSM-HardDeprecated78.7%
10
2024-06-20independentSource ↗
MT-BenchNearing Saturation8%
75
2024-06-19independentSource ↗
BIG-Bench HardSaturated82.3%
25
2024-06-19independentSource ↗
CodeXGLUEDeprecated71.4%
23
2024-06-19independentSource ↗
MBPP+Nearing Saturation77.5%
65
2024-06-18independentSource ↗
IFEvalDeprecated76.2%
25
2024-06-18independentSource ↗
WinoGrandeSaturated87.4%
28
2024-06-18independentSource ↗
GPQANearing Saturation74.4%
78
2024-06-18independentSource ↗
CommonsenseQASaturated88.2%
28
2024-06-17independentSource ↗
HumanEval+Saturated82.1%
25
2024-06-17independentSource ↗
GLUESaturated90.7%
28
2024-06-16independentSource ↗
MBPPSaturated80.6%
25
2024-06-16independentSource ↗
GSM8KSaturated88.6%
25
2024-06-16independentSource ↗
MMLUSaturated86.2%
25
2024-06-15independentSource ↗
HumanEvalSaturated86.6%
25
2024-06-15independentSource ↗
ARC-AGISaturated79%
25
2024-06-15independentSource ↗
Local Hardware Execution (Ollama)Hardware Compatible
# 1. Pull and execute model locally with Ollama
ollama run nemotron-4-340b-instruct

# 2. Or invoke via local OpenAI-compatible endpoint
curl http://localhost:11434/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "nemotron-4-340b-instruct",
    "messages": [{"role": "user", "content": "Analyze reasoning chains on AIME 2026"}]
  }'

Evaluation & Sourcing Notes

Scores listed for Nemotron-4-340B-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.