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. | 32k | 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. | 57.3% | 83.7% |
MATHReasoning — accuracy on the original competition mathematics test set | 27.9% | 64.3% |
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. | 55.7% | 84.7% |
DROPReasoning — three-shot token F1 on questions requiring reasoning over paragraphs | — | 83.2 |
GPQA DiamondScience — Graduate-level science questions in biology, physics, and chemistry — hard enough that subject-matter PhDs score around 65%. Higher is better. | — | 42.5% |
MGSMMultilingual math — solving grade-school word problems translated into multiple languages | — | 87.1% |
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. | 72% | — |
| Overview | ||
| Company | DeepSeek | OpenAI |
| Release date | May 16 2024 | Nov 6 2023 |
| Access | Open Weight | Closed |
| Model details | View model | View model |
Other comparisons
DeepSeek V2 LitevsClaude Fable 5.2GPT-4 TurbovsClaude Fable 5.2DeepSeek V2 LitevsGemini 4 HydrogenGPT-4 TurbovsGemini 4 HydrogenDeepSeek V2 LitevsMuse EmberGPT-4 TurbovsMuse EmberDeepSeek V2 LitevsGrok 4.7GPT-4 TurbovsGrok 4.7DeepSeek V2 LitevsMistral Large 4GPT-4 TurbovsMistral Large 4DeepSeek V2 LitevsKimi K3.1 ProGPT-4 TurbovsKimi K3.1 ProFrequently asked questions
GPT-4 Turbo leads DeepSeek V2 Lite on 3 of the 3 benchmarks they both report (HumanEval, MATH, MMLU). GPT-4 Turbo shipped 192 days before DeepSeek V2 Lite, so benchmark comparisons should account for the intervening progress.
Context windows are 32k (DeepSeek V2 Lite) vs 128k (GPT-4 Turbo). DeepSeek V2 Lite is open weight, while GPT-4 Turbo is closed.
On HumanEval, GPT-4 Turbo leads at 83.7% vs DeepSeek V2 Lite at 57.3%. On MATH, GPT-4 Turbo leads at 64.3% vs DeepSeek V2 Lite at 27.9%. On MMLU, GPT-4 Turbo leads at 84.7% vs DeepSeek V2 Lite at 55.7%.