Compare AI models
| Specifications | ||
ParametersThe learned values that shape how a model responds. B means billion and T means trillion. For mixture-of-experts models, the total count is larger than the number active for each token. Parameter count alone does not measure quality. | 16B | 51B |
Context windowHow much information the model can work with in one request, measured in tokens — pieces of words or other data. This includes your question, attachments, conversation and reply. Limits can vary by model variant and provider. | 8k | 128k |
| Benchmarks | ||
WinoGrandeReasoning — resolving ambiguous pronouns in context | 74.6% | 84.53% |
MMLUGeneral knowledge — A 57-subject multiple-choice exam — history, law, medicine, maths — that was the standard measure of how much a model knows from 2020 until roughly 2024, when frontier scores crowded into the high 80s and labs moved on to harder tests. The scores here were published years apart under different testing setups, so read them as a historical record rather than a like-for-like ranking. Higher is better. | 70% | 80.2% |
GSM8KGrade-school math — Grade-school maths word problems that take a few steps of arithmetic to work through. It separated the models of 2022 and 2023 sharply, then saturated. One caveat on the historical numbers: OpenAI included part of the GSM8K training set in GPT-4's pre-training mix, so GPT-4's score is not a clean few-shot result. Higher is better. | 77.4% | 91.43% |
LiveCodeBenchCompetitive coding — Coding problems published so recently the AI can't have seen them in training — a contamination-free test of raw programming skill. Higher is better. | 14.3% | — |
HumanEvalFunction synthesis — 164 small Python problems: the AI is given a function's description and has to write the working function. This was the coding benchmark of the GPT-3.5 and GPT-4 era, before the field moved to fixing real bugs in real repositories. Higher is better. | 48.1% | — |
MATHReasoning — accuracy on the original competition mathematics test set | 45.3% | — |
ARC-ChallengeReasoning — multiple-choice science questions from the AI2 Reasoning Challenge; unrelated to ARC-AGI | — | 69.2% |
HellaSwagReasoning — commonsense completion of short scenarios | — | 85.58% |
GPQA DiamondScience — Graduate-level science questions in biology, physics, and chemistry — hard enough that subject-matter PhDs score around 65%. Higher is better. | 35.2% | — |
SimpleQAKnowledge — accuracy on short, fact-seeking questions in the original SimpleQA set | 5.6% | — |
MMLU-ProKnowledge — accuracy on the ten-option MMLU-Pro benchmark | 42.4% | — |
| Overview | ||
| Company | Moonshot AI | NVIDIA |
| Release date | Feb 22 2025 | Sep 23 2024 |
| Access | Open Weight | Open Weight |
| Model details | View model | View model |
Other comparisons
MoonlightvsClaude Fable 5.2Llama 3.1 Nemotron 51BvsClaude Fable 5.2MoonlightvsGPT-6.2 AstraLlama 3.1 Nemotron 51BvsGPT-6.2 AstraMoonlightvsGemini 4 NeonLlama 3.1 Nemotron 51BvsGemini 4 NeonMoonlightvsMuse EmberLlama 3.1 Nemotron 51BvsMuse EmberMoonlightvsGrok 4.7Llama 3.1 Nemotron 51BvsGrok 4.7MoonlightvsDeepSeek-V4.1-Flash-1017Llama 3.1 Nemotron 51BvsDeepSeek-V4.1-Flash-1017Frequently asked questions
Llama 3.1 Nemotron 51B leads Moonlight on 3 of the 3 benchmarks they both report (WinoGrande, MMLU, GSM8K). Llama 3.1 Nemotron 51B shipped 152 days before Moonlight, so benchmark comparisons should account for the intervening progress.
Moonlight has 16B parameters, while Llama 3.1 Nemotron 51B has 51B. Context windows are 8k (Moonlight) vs 128k (Llama 3.1 Nemotron 51B).
On WinoGrande, Llama 3.1 Nemotron 51B leads at 84.53% vs Moonlight at 74.6%. On MMLU, Llama 3.1 Nemotron 51B leads at 80.2% vs Moonlight at 70%. On GSM8K, Llama 3.1 Nemotron 51B leads at 91.43% vs Moonlight at 77.4%.