DeepSeek-V3 Technical Report

By DeepSeek-AI, A. Liu, B. Feng, Bing Xue, Bing-Li Wang, Bochao Wu, Chengda Lu, Chenggang Zhao, C. Deng, Chenyu Zhang and 186 more. 2024

3,725 citations recorded by the source corpus.

Subfields: Large Language Models

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Large Language Models

1. Dahua Lin
2. Yinan He
3. Conghui He
4. Xizhou Zhu

Citations in the corpus

• Does Reinforcement Learning Really Incentivize Reasoning Capacity in LLMs Beyond the Base Model? (2025, neutral): “5-VL-7B (Bai et al., 2025) on Geometry3K (Lu et al., 2021), and evaluate its visual reasoning capabilities on filtered MathVista-TestMini (Lu et al., 2024) and MathVision-TestMini (Wang et al., 2024).”
• Visual-RFT: Visual Reinforcement Fine-Tuning (2025, neutral): “This integration enhances the models’ ability to understand complex multi-modal inputs and enables more advanced AI systems [15, 18, 40, 52] capable of processing and responding to both images and text.”
• Qwen3 Technical Report (2025, neutral): “5-72B-Base (Yang et al., 2024b), DeepSeek-V3 Base (Liu et al., 2024a).”
• VideoLLaMA 3: Frontier Multimodal Foundation Models for Image and Video Understanding (2025, neutral): “Recent years have witnessed the rapid growth of large language models (LLMs) [1–6], which significantly enhance natural language processing and understanding.”
• GLM-4.5: Agentic, Reasoning, and Coding (ARC) Foundation Models (2025, neutral): “We employ loss-free balance routing [40] and sigmoid gates for MoE layers [23].”
• Large Language Diffusion Models (2025, neutral): “Looking ahead, both the model scale and the amount of training data for LLaDA remain smaller than those of leading ARM counterparts [6, 26, 92–95], highlighting the need for further scaling to fully evaluate its capabilities.”
• FinGPT: Open-Source Financial Large Language Models (2023, neutral): “Large Language Models (LLMs) have been recognized as a technological breakthrough in NLP, such as GPT-3 and GPT-4 [ Brown et al. , 2020; Jiang et al. , 2023; OpenAI, 2023; Team et al. , 2023; Liu et al. , 2024 ] .”
• LLaVA-Prumerge: Adaptive Token Reduction for Efficient Large Multimodal Models (2024, neutral): “For example, the commonly used CLIP visual encoder only has 0.3B parameters, while the corresponding LLM such as LLaMA [43], Vicuna [45], DeepSeek [27], and Qwen [3] can have 7B, 13B, or even 70B+ parameters.”
• YOLOv12: Attention-Centric Real-Time Object Detectors (2025, neutral): “Interestingly, we find that this modification has only a slight impact on performance but significantly improves speed.”
• InternVL3.5: Advancing Open-Source Multimodal Models in Versatility, Reasoning, and Efficiency (2025, positive): “…to reduce padding tokens while balancing the token computation load across ranks for improved training efficiency, FP8 training based on DeepGEMM [68] and liger-kernel’s fused cross-entropy operator [47] to accelerate the training process, FlashAttention-3 [26, 27] to support packed inputs and…”
• Group-in-Group Policy Optimization for LLM Agent Training (2025, neutral): “Large Language Models (LLMs) [1, 2, 3, 4] have leapt from static question-answer systems to versatile agents that perceive, reason, and act in open-ended environments.”
• DAPO: An Open-Source LLM Reinforcement Learning System at Scale (2025, neutral): “Test-time scaling such as OpenAI’s o1 [1] and DeepSeek’s R1 [2] brings a profound paradigm shift to Large Language Models (LLMs) [3–7].”
• Understanding R1-Zero-Like Training: A Critical Perspective (2025, neutral): “4 where DeepSeek-V3-Base generates keywords such as “Aha”, “wait”, and “verify the problem”.”
• ReMindRAG: Low-Cost LLM-Guided Knowledge Graph Traversal for Efficient RAG (2025, unknown): “In our experiments, we employed GPT-4o-mini [34] and Deepseek-V3 [28] as the underlying large language models (LLMs).”