NVIDIA AI & NIMopen-weight
Released: 2024-06-14Nemotron-4-340B-Reward
Top-performing reward model on RewardBench for scoring alignment, safety, helpfulness, and response quality in RLHF loops.
Intelligence Index
86.0/ 100
Calibrated multi-domain compositeThroughput Speed
45tok/sec
Output streaming throughputTime To First Token
480ms
Initial chunk server latencyLiveBench Contam-Free
74.5%
September 2026 suite scoreSourced Benchmark Evaluations (51)
Verified performance across authoritative benchmarks with provenance tracking.| Benchmark | Status | Score | Trust | Date | Source Type | Provenance |
|---|---|---|---|---|---|---|
| LiveBench (Continuous) | Active | 74.5% | 94 | 2025-02-01 | independent | Source ↗ |
| ARC-AGI-3 | Active | 55.3% | 93 | 2024-07-19 | independent | Source ↗ |
| ARC-AGI-2 | Nearing Saturation | 66.3% | 75 | 2024-07-12 | independent | Source ↗ |
| FrontierMath | Active | 23.5% | 98 | 2024-07-09 | independent | Source ↗ |
| PutnamBench | Active | 43.7% | 90 | 2024-07-08 | independent | Source ↗ |
| SWE-Lancer | Deprecated | 53% | 20 | 2024-07-06 | independent | Source ↗ |
| Omni-MATH | Deprecated | 62.2% | 10 | 2024-07-06 | independent | Source ↗ |
| Soohak | Active | 63.8% | 98 | 2024-07-04 | independent | Source ↗ |
| FrontierCode | Active | 45.2% | 98 | 2024-07-04 | independent | Source ↗ |
| AIME | Nearing Saturation | 79.8% | 75 | 2024-07-02 | independent | Source ↗ |
| StructFlowBench | Nearing Saturation | 67.3% | 75 | 2024-07-02 | independent | Source ↗ |
| Humanity's Last Exam | Active | 37.4% | 93 | 2024-07-02 | independent | Source ↗ |
| OlympiadBench | Active | 68.9% | 80 | 2024-07-01 | independent | Source ↗ |
| MathArena | Active | 67.2% | 99 | 2024-06-30 | independent | Source ↗ |
| Terminal-Bench | Active | 54.6% | 80 | 2024-06-30 | independent | Source ↗ |
| FireBench | Active | 66.5% | 95 | 2024-06-30 | independent | Source ↗ |
| SimpleQA | Active | 44.7% | 90 | 2024-06-29 | independent | Source ↗ |
| CodeContests | Deprecated | 50.7% | 10 | 2024-06-29 | independent | Source ↗ |
| ComplexBench | Nearing Saturation | 64.9% | 80 | 2024-06-28 | independent | Source ↗ |
| SWE-bench Pro | Active | 48.4% | 90 | 2024-06-28 | independent | Source ↗ |
| MultiPL-E | Deprecated | 71.8% | 10 | 2024-06-27 | independent | Source ↗ |
| LiveCodeBench | Saturated | 64% | 39 | 2024-06-26 | independent | Source ↗ |
| MATH | Saturated | 82.3% | 25 | 2024-06-26 | independent | Source ↗ |
| AgentIF | Nearing Saturation | 53.8% | 80 | 2024-06-26 | independent | Source ↗ |
| Mercury | Deprecated | 68.6% | 10 | 2024-06-25 | independent | Source ↗ |
| MMLU-Pro | Nearing Saturation | 79.1% | 75 | 2024-06-25 | independent | Source ↗ |
| CyberSecEval | Deprecated | 70.2% | 10 | 2024-06-24 | independent | Source ↗ |
| TabMWP | Saturated | 77.3% | 25 | 2024-06-24 | independent | Source ↗ |
| SWE-bench Verified | Saturated | 60.8% | 25 | 2024-06-23 | independent | Source ↗ |
| MMLU-Redux | Saturated | 84.3% | 25 | 2024-06-23 | independent | Source ↗ |
| SWE-bench | Deprecated | 53% | 10 | 2024-06-22 | independent | Source ↗ |
| MGSM | Nearing Saturation | 79.8% | 65 | 2024-06-22 | independent | Source ↗ |
| TruthfulQA | Saturated | 75.7% | 25 | 2024-06-21 | independent | Source ↗ |
| AGIEval | Saturated | 73.1% | 25 | 2024-06-21 | independent | Source ↗ |
| SuperGLUE | Saturated | 86.7% | 25 | 2024-06-20 | independent | Source ↗ |
| GSM-Hard | Deprecated | 80.6% | 10 | 2024-06-20 | independent | Source ↗ |
| MT-Bench | Nearing Saturation | 8.1% | 75 | 2024-06-19 | independent | Source ↗ |
| BIG-Bench Hard | Saturated | 83.3% | 25 | 2024-06-19 | independent | Source ↗ |
| CodeXGLUE | Deprecated | 73.3% | 23 | 2024-06-19 | independent | Source ↗ |
| MBPP+ | Nearing Saturation | 79.6% | 65 | 2024-06-18 | independent | Source ↗ |
| IFEval | Deprecated | 77.5% | 25 | 2024-06-18 | independent | Source ↗ |
| WinoGrande | Saturated | 88.4% | 28 | 2024-06-18 | independent | Source ↗ |
| GPQA | Nearing Saturation | 75.7% | 78 | 2024-06-18 | independent | Source ↗ |
| CommonsenseQA | Saturated | 89.3% | 28 | 2024-06-17 | independent | Source ↗ |
| HumanEval+ | Saturated | 84.2% | 25 | 2024-06-17 | independent | Source ↗ |
| GLUE | Saturated | 91.8% | 28 | 2024-06-16 | independent | Source ↗ |
| MBPP | Saturated | 82.7% | 25 | 2024-06-16 | independent | Source ↗ |
| GSM8K | Saturated | 90.7% | 25 | 2024-06-16 | independent | Source ↗ |
| MMLU | Saturated | 87.7% | 25 | 2024-06-15 | independent | Source ↗ |
| HumanEval | Saturated | 88.9% | 25 | 2024-06-15 | independent | Source ↗ |
| ARC-AGI | Saturated | 79.9% | 25 | 2024-06-15 | independent | Source ↗ |
Local Hardware Execution (Ollama)Hardware Compatible
# 1. Pull and execute model locally with Ollama
ollama run nemotron-4-340b-reward
# 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-reward",
"messages": [{"role": "user", "content": "Analyze reasoning chains on AIME 2026"}]
}'Evaluation & Sourcing Notes
Scores listed for Nemotron-4-340B-Reward 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.