DeepSeek V2 LiteOpen Weight

Released

DeepSeek V2 Lite is an AI model released by DeepSeek on Thursday, May 16 2024, 10 days after DeepSeek V2. It is an open-weight model — the trained weights are available to download and run. It is a 16B parameterThe 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. model with a 32k token 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.. Benchmark results (shown below) cover HumanEval, MATH, MMLU, and GSM8K.

Benchmarks

Coding

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%
#53 of 65

Reasoning & science

MATHReasoning — accuracy on the original competition mathematics test set
27.9%
#36 of 43
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%
#58 of 59
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%
#24 of 31

Compare DeepSeek V2 Lite with

DeepSeek V2 Lite

About

DeepSeek V2 Lite followed the original V2 release on May 16, 2024. It brought Multi-head Latent Attention and sparse expert routing to a smaller model intended to make the architecture easier to study and deploy.

The release included base and chat versions with a 32K context window. The scores shown here use the chat checkpoint; they should not be read as results for the larger 236B V2 model.

Frequently asked questions

DeepSeek V2 Lite was released by DeepSeek on Thursday, May 16 2024.

All DeepSeek releases

30 tracked