GitHub – Deepseek-ai/DeepSeek-V3

We provide DeepSeek-V3, a strong Mixture-of-Experts (MoE) language design with 671B total parameters with 37B activated for each token. To attain effective reasoning and cost-effective training, DeepSeek-V3 embraces Multi-head Latent Attention (MLA) and DeepSeekMoE architectures, which were completely confirmed in DeepSeek-V2. Furthermore, DeepSeek-V3 leaders an auxiliary-loss-free technique for load balancing and sets a multi-token forecast training objective for more powerful performance. We pre-train DeepSeek-V3 on 14.8 trillion varied and high-quality tokens, followed by Supervised Fine-Tuning and Reinforcement Learning stages to fully harness its abilities. Comprehensive evaluations expose that DeepSeek-V3 surpasses other open-source designs and attains performance equivalent to leading closed-source designs. Despite its outstanding performance, DeepSeek-V3 requires just 2.788 M H800 GPU hours for its full training. In addition, its training process is incredibly stable. Throughout the whole procedure, we did not experience any irrecoverable loss spikes or perform any rollbacks.

2. Model Summary

Architecture: Innovative Load Balancing Strategy and Training Objective

– On top of the efficient architecture of DeepSeek-V2, we leader an auxiliary-loss-free strategy for load balancing, which minimizes the performance destruction that emerges from encouraging load balancing.
– We investigate a Multi-Token Prediction (MTP) objective and show it useful to model performance. It can also be utilized for speculative decoding for reasoning acceleration.

Pre-Training: Towards Ultimate Training Efficiency

– We design an FP8 mixed precision training structure and, for the very first time, verify the expediency and effectiveness of FP8 training on an incredibly large-scale design.
– Through co-design of algorithms, frameworks, and hardware, we overcome the communication traffic jam in cross-node MoE training, almost attaining full computation-communication overlap.
This substantially boosts our training efficiency and reduces the training expenses, allowing us to even more scale up the design size without extra overhead.
– At an economical cost of only 2.664 M H800 GPU hours, we complete the pre-training of DeepSeek-V3 on 14.8 T tokens, producing the currently strongest open-source base model. The subsequent training stages after pre-training require just 0.1 M GPU hours.

Post-Training: Knowledge Distillation from DeepSeek-R1

– We introduce an innovative method to boil down thinking abilities from the long-Chain-of-Thought (CoT) design, particularly from among the DeepSeek R1 series designs, into standard LLMs, particularly DeepSeek-V3. Our pipeline elegantly incorporates the confirmation and reflection patterns of R1 into DeepSeek-V3 and especially enhances its reasoning performance. Meanwhile, we also preserve a control over the output design and length of DeepSeek-V3.

3. Model Downloads

The overall size of DeepSeek-V3 designs on Hugging Face is 685B, which consists of 671B of the Main Model weights and 14B of the Multi-Token Prediction (MTP) Module weights. **

To make sure optimal efficiency and flexibility, we have partnered with open-source communities and hardware vendors to provide numerous methods to run the model in your area. For step-by-step assistance, have a look at Section 6: How_to Run_Locally.

For developers wanting to dive deeper, we recommend checking out README_WEIGHTS. md for information on the Main Model weights and the Multi-Token Prediction (MTP) Modules. Please note that MTP assistance is presently under active development within the community, and we invite your contributions and feedback.

4. Evaluation Results

Base Model

Standard Benchmarks

Best outcomes are revealed in vibrant. Scores with a space not going beyond 0.3 are thought about to be at the same level. DeepSeek-V3 attains the very best efficiency on the majority of standards, particularly on math and code jobs. For more evaluation details, please check our paper.

Context Window

Evaluation results on the Needle In A Haystack (NIAH) tests. DeepSeek-V3 performs well across all context window lengths up to 128K.

Chat Model

Standard Benchmarks (Models bigger than 67B)

All designs are assessed in a setup that restricts the output length to 8K. Benchmarks including fewer than 1000 samples are checked several times utilizing differing temperature level settings to obtain robust outcomes. DeepSeek-V3 stands as the best-performing open-source model, and also exhibits competitive efficiency against frontier closed-source models.

Open Ended Generation Evaluation

English open-ended discussion examinations. For AlpacaEval 2.0, we use the length-controlled win rate as the metric.

5. Chat Website & API Platform

You can chat with DeepSeek-V3 on DeepSeek’s main website: chat.deepseek.com

We also offer OpenAI-Compatible API at DeepSeek Platform: platform.deepseek.com

6. How to Run Locally

DeepSeek-V3 can be released locally utilizing the following hardware and open-source community software:

DeepSeek-Infer Demo: We offer a simple and lightweight demo for FP8 and BF16 inference.
SGLang: Fully support the DeepSeek-V3 model in both BF16 and FP8 inference modes, with Multi-Token Prediction coming soon.
LMDeploy: Enables effective FP8 and BF16 reasoning for local and cloud implementation.
TensorRT-LLM: Currently supports BF16 reasoning and INT4/8 quantization, with FP8 assistance coming quickly.
vLLM: Support DeepSeek-V3 design with FP8 and BF16 modes for tensor parallelism and pipeline parallelism.
AMD GPU: Enables running the DeepSeek-V3 model on AMD GPUs through SGLang in both BF16 and FP8 modes.
Huawei Ascend NPU: Supports running DeepSeek-V3 on Huawei Ascend devices.
Since FP8 training is natively adopted in our structure, we just provide FP8 weights. If you need BF16 weights for experimentation, you can use the provided conversion script to perform the change.

Here is an example of converting FP8 weights to BF16:

Hugging Face’s Transformers has not been directly supported yet. **

6.1 Inference with DeepSeek-Infer Demo (example just)

System Requirements

Note

Linux with Python 3.10 only. Mac and Windows are not supported.

Dependencies:

Model Weights & Demo Code Preparation

First, clone our DeepSeek-V3 GitHub repository:

Navigate to the reasoning folder and set up dependences noted in requirements.txt. Easiest way is to use a bundle supervisor like conda or uv to produce a new virtual environment and install the dependences.

Download the design weights from Hugging Face, and put them into/ path/to/DeepSeek-V 3 folder.

Model Weights Conversion

Convert Hugging Face design weights to a particular format:

Run

Then you can talk with DeepSeek-V3:

Or batch reasoning on a provided file:

6.2 Inference with SGLang (recommended)

SGLang presently supports MLA optimizations, DP Attention, FP8 (W8A8), FP8 KV Cache, and Torch Compile, delivering state-of-the-art latency and throughput performance among open-source structures.

Notably, SGLang v0.4.1 fully supports running DeepSeek-V3 on both NVIDIA and AMD GPUs, making it a highly flexible and robust service.

SGLang also supports multi-node tensor parallelism, allowing you to run this design on several network-connected machines.

Multi-Token Prediction (MTP) is in development, and progress can be tracked in the optimization plan.

Here are the launch instructions from the SGLang team: https://github.com/sgl-project/sglang/tree/main/benchmark/deepseek_v3

6.3 Inference with LMDeploy (suggested)

LMDeploy, a versatile and high-performance reasoning and serving structure tailored for large language designs, now supports DeepSeek-V3. It offers both offline pipeline processing and online implementation capabilities, perfectly incorporating with PyTorch-based workflows.

For extensive detailed instructions on running DeepSeek-V3 with LMDeploy, please describe here: InternLM/lmdeploy # 2960

6.4 Inference with TRT-LLM (recommended)

TensorRT-LLM now supports the DeepSeek-V3 design, using accuracy choices such as BF16 and INT4/INT8 weight-only. Support for FP8 is presently in progress and will be launched soon. You can access the custom branch of TRTLLM particularly for DeepSeek-V3 support through the following link to experience the brand-new functions straight: https://github.com/NVIDIA/TensorRT-LLM/tree/deepseek/examples/deepseek_v3.

6.5 Inference with vLLM (suggested)

vLLM v0.6.6 supports DeepSeek-V3 reasoning for FP8 and BF16 modes on both NVIDIA and AMD GPUs. Aside from standard methods, vLLM uses pipeline parallelism allowing you to run this model on numerous devices linked by networks. For in-depth guidance, please describe the vLLM directions. Please feel totally free to follow the enhancement strategy also.

6.6 Recommended Inference Functionality with AMD GPUs

In cooperation with the AMD team, we have accomplished Day-One support for AMD GPUs using SGLang, with complete compatibility for both FP8 and BF16 precision. For detailed guidance, please describe the SGLang guidelines.

6.7 Recommended Inference Functionality with Huawei Ascend NPUs

The MindIE structure from the Huawei Ascend community has effectively adapted the BF16 version of DeepSeek-V3. For step-by-step guidance on Ascend NPUs, please follow the instructions here.

7. License

This code repository is accredited under the MIT License. The use of DeepSeek-V3 Base/Chat models goes through the Model License. DeepSeek-V3 series (consisting of Base and Chat) supports industrial use.