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 | — |
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 | ||
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% | 92.4% |
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% | 73.3% |
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% | 90.8% |
SWE-Bench VerifiedCoding — Real coding tasks pulled from open-source projects — the AI has to find and fix actual bugs. A human-checked version of the original SWE-Bench. Higher is better. | — | 41.4% |
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% | — |
MATHReasoning — accuracy on the original competition mathematics test set | 45.3% | — |
MATH-500Reasoning — accuracy on the 500-problem MATH subset | — | 85.5% |
DROPReasoning — three-shot token F1 on questions requiring reasoning over paragraphs | — | 74.8 |
WinoGrandeReasoning — resolving ambiguous pronouns in context | 74.6% | — |
AIME 2024Reasoning — single-answer accuracy on the 2024 invitational mathematics exam without tools | — | 44.6% |
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% | — |
MGSMMultilingual math — solving grade-school word problems translated into multiple languages | — | 90.8% |
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% | — |
| Overview | ||
| Company | Moonshot AI | OpenAI |
| Release date | Feb 22 2025 | Sep 12 2024 |
| Access | Open Weight | Closed |
| Model details | View model | View model |
Other comparisons
MoonlightvsClaude Fable 5.2o1-previewvsClaude Fable 5.2MoonlightvsGemini 4 Hydrogeno1-previewvsGemini 4 HydrogenMoonlightvsMuse Embero1-previewvsMuse EmberMoonlightvsGrok 4.7o1-previewvsGrok 4.7MoonlightvsDeepSeek-V4.1-Flash-1017o1-previewvsDeepSeek-V4.1-Flash-1017MoonlightvsMistral Large 4o1-previewvsMistral Large 4Frequently asked questions
o1-preview leads Moonlight on 3 of the 3 benchmarks they both report (HumanEval, GPQA Diamond, MMLU). o1-preview shipped 163 days before Moonlight, so benchmark comparisons should account for the intervening progress.
Context windows are 8k (Moonlight) vs 128k (o1-preview). Moonlight is open weight, while o1-preview is closed.
On HumanEval, o1-preview leads at 92.4% vs Moonlight at 48.1%. On GPQA Diamond, o1-preview leads at 73.3% vs Moonlight at 35.2%. On MMLU, o1-preview leads at 90.8% vs Moonlight at 70%.