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. | 671B | — |
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. | 128k | 200k |
| Benchmarks | ||
MATH-500Reasoning — accuracy on the 500-problem MATH subset | 97.3% | 98.2% |
DROPReasoning — three-shot token F1 on questions requiring reasoning over paragraphs | 92.2 | 78.1 |
GPQA DiamondScience — Graduate-level science questions in biology, physics, and chemistry — hard enough that subject-matter PhDs score around 65%. Higher is better. | 71.5% | 81.3% |
SimpleQAKnowledge — accuracy on short, fact-seeking questions in the original SimpleQA set | 30.1% | 19.3% |
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. | 90.8% | 90.3% |
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. | 49.2% | — |
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. | 65.9% | — |
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. | — | 99.3% |
AIME 2024Reasoning — single-answer accuracy on the 2024 invitational mathematics exam without tools | 79.8% | — |
MMLU-ProKnowledge — accuracy on the ten-option MMLU-Pro benchmark | 84% | — |
MGSMMultilingual math — solving grade-school word problems translated into multiple languages | — | 93.5% |
| Overview | ||
| Company | DeepSeek | OpenAI |
| Release date | Jan 20 2025 | Apr 16 2025 |
| Access | Open Weight | Closed |
| Model details | View model | View model |
Other comparisons
DeepSeek R1vsClaude Fable 5.2o4-mini-highvsClaude Fable 5.2DeepSeek R1vsGemini 4 Hydrogeno4-mini-highvsGemini 4 HydrogenDeepSeek R1vsMuse Embero4-mini-highvsMuse EmberDeepSeek R1vsGrok 4.7o4-mini-highvsGrok 4.7DeepSeek R1vsMistral Large 4o4-mini-highvsMistral Large 4DeepSeek R1vsKimi K3.1 Proo4-mini-highvsKimi K3.1 ProFrequently asked questions
DeepSeek R1 leads o4-mini-high on 3 of the 5 benchmarks they both report. DeepSeek R1 shipped 86 days before o4-mini-high, so benchmark comparisons should account for the intervening progress.
Context windows are 128k (DeepSeek R1) vs 200k (o4-mini-high). DeepSeek R1 is open weight, while o4-mini-high is closed.
On MATH-500, o4-mini-high leads at 98.2% vs DeepSeek R1 at 97.3%. On DROP, DeepSeek R1 leads at 92.2 vs o4-mini-high at 78.1. On GPQA Diamond, o4-mini-high leads at 81.3% vs DeepSeek R1 at 71.5%. On SimpleQA, DeepSeek R1 leads at 30.1% vs o4-mini-high at 19.3%. On MMLU, DeepSeek R1 leads at 90.8% vs o4-mini-high at 90.3%.