entity

DeepSeek-AI

Also known as: DeepSeek AI, DeepSeek-AI, DeepSeek

Facts (18)

Sources

The Impact of Open Source on Digital Innovation linkedin.com LinkedIn 7 facts

measurementDeepSeek built its AI model in a period of 2 months.

claimThe success of the DeepSeek AI model is built upon open source software, hardware, and open research, much of which was originally developed in the United States.

measurementDeepSeek's AI model costs 10 cents per million tokens, whereas GPT-4 costs $4.40 per million tokens.

claimDeepSeek's AI model utilizes lower-end GPUs for operation.

measurementDeepSeek spent $5.6 million to build its AI model, compared to OpenAI's reported $5 billion per year expenditure.

claimDeepSeek, a Chinese AI lab, developed an AI model that matches or outperforms GPT-4 in several benchmarks.

claimDeepSeek released code and model weights back to the open source community, allowing these resources to be leveraged for further innovations globally, including in the United States.

A Comprehensive Benchmark and Evaluation Framework for Multi ... arxiv.org arXiv Jan 6, 2026 3 facts

measurementThe Majority Voting strategy for ensemble LLM judges consistently produces stable agreement with human clinical experts, maintaining F1-scores in the 75–79% range across Doctor Agents including DeepSeek, Gemini, and GPT-5.

referenceDeepSeek-AI published the DeepSeek-R1 technical report in 2025, detailing the use of reinforcement learning to incentivize reasoning capabilities in large language models.

referenceDeepSeek-AI published the DeepSeek-V3 technical report in 2024.

Survey and analysis of hallucinations in large language models frontiersin.org Frontiers Sep 29, 2025 2 facts

referenceThe study selected the following open-source Large Language Models (LLMs) for evaluation: LLaMA 2 (13B) (Meta AI, 2023), a transformer-based model fine-tuned for dialogue; Mistral 7B instruct, an instruction-tuned model; DeepSeek 67B (DeepSeek AI, 2023), a multilingual model trained on code and web data; OpenChat-3.5 (Openchat Team, 2023), a community-finetuned model derived from LLaMA; and Gwen, an open-access research model emphasizing retrieval-enhanced factual generation.

claimLarge Language Models including GPT-3 (Brown et al., 2020), GPT-4 (OpenAI, 2023b), LLaMA 2 (Touvron et al., 2023), Claude (Anthropic, 2023), and DeepSeek (DeepSeek AI, 2023) have demonstrated capabilities in zero-shot and few-shot learning tasks.

Strategic Decoupling and Its Implications for US-China Relations rsis.edu.sg RSIS Sep 1, 2025 2 facts

claimChinese firms, including DeepSeek, achieved breakthroughs in AI, robotics, pharmaceuticals, and defense technology by late 2024.

claimBy late 2024, Chinese firms such as DeepSeek achieved notable breakthroughs in AI, robotics, pharmaceuticals, and defence technology.

Media Coverage - News Center - Baruch College newscenter.baruch.cuny.edu Baruch College 1 fact

claimNizan Geslevich Packin analyzed privacy concerns and government actions regarding TikTok and DeepSeek in China.

What Is Open Source Software? - IBM ibm.com IBM 1 fact

measurementIn 2025, the Chinese company DeepSeek released R1, a large language model that cost USD 5.6 million to train.

Policymakers Overlook How Open Source AI Is Reshaping ... techpolicy.press Lucie-Aimée Kaffee, Shayne Longpre · Tech Policy Press Dec 9, 2025 1 fact

accountSince early 2025, China's presence in the open-source AI ecosystem has expanded rapidly, driven by the ascent of companies such as DeepSeek and Alibaba, whose models have become global defaults.

The U.S.-China Trade Relationship | Council on Foreign Relations cfr.org Council on Foreign Relations Oct 31, 2025 1 fact

accountIn January 2025, the Chinese startup DeepSeek launched an advanced AI model that operates at lower costs and higher energy efficiency, rivaling the capacity of U.S. AI companies like OpenAI and Google DeepMind.