We are excited to announce the official release of DeepSeek-V3.2-Exp, an experimental version of our model. As an intermediate step toward our next-generation architecture, V3.2-Exp builds upon V3.1-Terminus by introducing DeepSeek Sparse Attention—a sparse attention mechanism designed to explore and validate optimizations for training and inference efficiency in long-context scenarios.
This experimental release represents our ongoing research into more efficient transformer architectures, particularly focusing on improving computational efficiency when processing extended text sequences.
DeepSeek Sparse Attention (DSA) achieves fine-grained sparse attention for the first time, delivering substantial improvements in long-context training and inference efficiency while maintaining virtually identical model output quality.
To rigorously evaluate the impact of introducing sparse attention, we deliberately aligned the training configurations of DeepSeek-V3.2-Exp with V3.1-Terminus. Across public benchmarks in various domains, DeepSeek-V3.2-Exp demonstrates performance on par with V3.1-Terminus.
Benchmark
DeepSeek-V3.1-Terminus
DeepSeek-V3.2-Exp
Reasoning Mode w/o Tool Use
MMLU-Pro
85.0
85.0
GPQA-Diamond
80.7
79.9
Humanity's Last Exam
21.7
19.8
LiveCodeBench
74.9
74.1
AIME 2025
88.4
89.3
HMMT 2025
86.1
83.6
Codeforces
2046
2121
Aider-Polyglot
76.1
74.5
Agentic Tool Use
BrowseComp
38.5
40.1
BrowseComp-zh
45.0
47.9
SimpleQA
96.8
97.1
SWE Verified
68.4
67.8
SWE-bench Multilingual
57.8
57.9
Terminal-bench
36.7
37.7
How to Run Locally
HuggingFace
We provide an updated inference demo code in the
inference
folder to help the community quickly get started with our model and understand its architectural details.
First convert huggingface model weights to the the format required by our inference demo. Set
MP
to match your available GPU count:
vLLM provides day-0 support of DeepSeek-V3.2-Exp. See the
recipes
for up-to-date details.
Open-Source Kernels
For TileLang kernels with
better readability and research-purpose design
, please refer to
TileLang
.
For
high-performance CUDA kernels
, indexer logit kernels (including paged versions) are available in
DeepGEMM
. Sparse attention kernels are released in
FlashMLA
.
License
This repository and the model weights are licensed under the
MIT License
.
DeepSeek-V3.2-Exp huggingface.co is an AI model on huggingface.co that provides DeepSeek-V3.2-Exp's model effect (), which can be used instantly with this deepseek-ai DeepSeek-V3.2-Exp model. huggingface.co supports a free trial of the DeepSeek-V3.2-Exp model, and also provides paid use of the DeepSeek-V3.2-Exp. Support call DeepSeek-V3.2-Exp model through api, including Node.js, Python, http.
DeepSeek-V3.2-Exp huggingface.co is an online trial and call api platform, which integrates DeepSeek-V3.2-Exp's modeling effects, including api services, and provides a free online trial of DeepSeek-V3.2-Exp, you can try DeepSeek-V3.2-Exp online for free by clicking the link below.
deepseek-ai DeepSeek-V3.2-Exp online free url in huggingface.co:
DeepSeek-V3.2-Exp is an open source model from GitHub that offers a free installation service, and any user can find DeepSeek-V3.2-Exp on GitHub to install. At the same time, huggingface.co provides the effect of DeepSeek-V3.2-Exp install, users can directly use DeepSeek-V3.2-Exp installed effect in huggingface.co for debugging and trial. It also supports api for free installation.