GitHub – Deepseek-ai/DeepSeek-V3

We present DeepSeek-V3, a strong Mixture-of-Experts (MoE) language model with 671B total specifications with 37B triggered for each token. To achieve efficient reasoning and affordable training, DeepSeek-V3 adopts Multi-head Latent Attention (MLA) and DeepSeekMoE architectures, which were thoroughly verified in DeepSeek-V2. Furthermore, DeepSeek-V3 leaders an auxiliary-loss-free strategy 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 premium tokens, followed by Supervised Fine-Tuning and Reinforcement Learning phases to fully harness its abilities. Comprehensive evaluations reveal that DeepSeek-V3 surpasses other open-source models and accomplishes efficiency similar to leading closed-source models. Despite its excellent efficiency, DeepSeek-V3 requires only 2.788 M H800 GPU hours for its full training. In addition, its training process is incredibly stable. Throughout the whole training 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 effective architecture of DeepSeek-V2, we pioneer an auxiliary-loss-free technique for load balancing, which reduces the performance degradation that emerges from motivating load balancing.
– We investigate a Multi-Token Prediction (MTP) objective and prove it useful to model performance. It can likewise be used for speculative decoding for reasoning acceleration.

Pre-Training: Towards Ultimate Training Efficiency

– We develop an FP8 mixed precision training framework and, for the very first time, confirm the expediency and effectiveness of FP8 training on a very large-scale model.
– Through co-design of algorithms, frameworks, and hardware, we conquer the communication bottleneck in cross-node MoE training, nearly attaining full computation-communication overlap.
This significantly boosts our training effectiveness and lowers the training expenses, allowing us to further scale up the model size without extra overhead.
– At an affordable cost of just 2.664 M H800 GPU hours, we finish the pre-training of DeepSeek-V3 on 14.8 T tokens, producing the currently base design. The subsequent training phases after pre-training need only 0.1 M GPU hours.

Post-Training: Knowledge Distillation from DeepSeek-R1

– We present an ingenious method to boil down reasoning capabilities from the long-Chain-of-Thought (CoT) model, particularly from one of the DeepSeek R1 series models, into standard LLMs, especially DeepSeek-V3. Our pipeline elegantly incorporates the confirmation and reflection patterns of R1 into DeepSeek-V3 and especially enhances its reasoning efficiency. Meanwhile, we also keep a control over the output design and length of DeepSeek-V3.

3. Model Downloads

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

To make sure optimal performance and versatility, we have partnered with open-source neighborhoods and hardware suppliers to supply several methods to run the model in your area. For step-by-step guidance, have a look at Section 6: How_to Run_Locally.

For designers seeking to dive much deeper, we suggest exploring README_WEIGHTS. md for information on the Main Model weights and the Multi-Token Prediction (MTP) Modules. Please note that MTP assistance is currently under active development within the community, and we invite your contributions and feedback.

4. Evaluation Results

Base Model

Standard Benchmarks

Best outcomes are displayed in bold. Scores with a space not going beyond 0.3 are considered to be at the very same level. DeepSeek-V3 attains the best performance on many criteria, particularly on mathematics and code jobs. For more assessment details, please check our paper.

Context Window

Evaluation results on the Needle In A Haystack (NIAH) tests. DeepSeek-V3 carries out well across all context window lengths as much as 128K.

Chat Model

Standard Benchmarks (Models larger than 67B)

All models are evaluated in a configuration that limits the output length to 8K. Benchmarks consisting of fewer than 1000 samples are tested several times utilizing differing temperature level settings to derive robust outcomes. DeepSeek-V3 stands as the best-performing open-source model, and also shows competitive efficiency versus frontier closed-source designs.

Open Ended Generation Evaluation

English open-ended conversation assessments. For AlpacaEval 2.0, we utilize 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 provide 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 provide a basic and lightweight demonstration for FP8 and BF16 inference.
SGLang: Fully support the DeepSeek-V3 model in both BF16 and FP8 reasoning modes, with Multi-Token Prediction coming quickly.
LMDeploy: Enables efficient FP8 and BF16 inference for local and cloud implementation.
TensorRT-LLM: Currently supports BF16 reasoning and INT4/8 quantization, with FP8 assistance coming soon.
vLLM: Support DeepSeek-V3 design with FP8 and BF16 modes for tensor parallelism and pipeline parallelism.
AMD GPU: Enables running the DeepSeek-V3 design on AMD GPUs via SGLang in both BF16 and FP8 modes.
Huawei Ascend NPU: Supports running DeepSeek-V3 on Huawei Ascend devices.
Since FP8 training is natively embraced in our framework, we just supply FP8 weights. If you need BF16 weights for experimentation, you can utilize the offered conversion script to perform the improvement.

Here is an example of converting FP8 weights to BF16:

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

6.1 Inference with DeepSeek-Infer Demo (example only)

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 inference folder and install dependences noted in requirements.txt. Easiest method is to utilize a bundle manager like conda or uv to create a brand-new virtual environment and set up the dependencies.

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 specific format:

Run

Then you can talk with DeepSeek-V3:

Or batch reasoning on a provided file:

6.2 Inference with SGLang (recommended)

SGLang currently supports MLA optimizations, DP Attention, FP8 (W8A8), FP8 KV Cache, and Torch Compile, providing state-of-the-art latency and throughput efficiency among open-source frameworks.

Notably, SGLang v0.4.1 completely supports running DeepSeek-V3 on both NVIDIA and AMD GPUs, making it an extremely versatile and robust option.

SGLang likewise supports multi-node tensor parallelism, allowing you to run this model on several network-connected makers.

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

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 (recommended)

LMDeploy, a versatile and high-performance reasoning and serving framework tailored for large language models, now supports DeepSeek-V3. It uses both offline pipeline processing and online release abilities, flawlessly integrating with PyTorch-based workflows.

For detailed step-by-step directions on running DeepSeek-V3 with LMDeploy, please refer to here: InternLM/lmdeploy # 2960

6.4 Inference with TRT-LLM (recommended)

TensorRT-LLM now supports the DeepSeek-V3 design, providing precision options such as BF16 and INT4/INT8 weight-only. Support for FP8 is currently in development and will be released soon. You can access the custom-made branch of TRTLLM specifically for DeepSeek-V3 support through the following link to experience the new functions directly: https://github.com/NVIDIA/TensorRT-LLM/tree/deepseek/examples/deepseek_v3.

6.5 Inference with vLLM (recommended)

vLLM v0.6.6 supports DeepSeek-V3 reasoning for FP8 and BF16 modes on both NVIDIA and AMD GPUs. Aside from standard methods, vLLM provides pipeline parallelism allowing you to run this model on multiple machines linked by networks. For in-depth assistance, please refer to the vLLM instructions. Please do not hesitate to follow the improvement plan too.

6.6 Recommended Inference Functionality with AMD GPUs

In partnership with the AMD team, we have attained Day-One support for AMD GPUs utilizing SGLang, with complete compatibility for both FP8 and BF16 accuracy. For comprehensive guidance, please refer to the SGLang instructions.

6.7 Recommended Inference Functionality with Huawei Ascend NPUs

The MindIE framework from the Huawei Ascend neighborhood has successfully adjusted the BF16 variation of DeepSeek-V3. For detailed guidance on Ascend NPUs, please follow the instructions here.

7. License

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