Close Menu
    Facebook LinkedIn YouTube Instagram X (Twitter)
    Blue Tech Wave Media
    Facebook LinkedIn YouTube Instagram X (Twitter)
    • Home
    • Leadership Alliance
    • Exclusives
    • Internet Governance
      • Regulation
      • Governance Bodies
      • Emerging Tech
    • IT Infrastructure
      • Networking
      • Cloud
      • Data Centres
    • Company Stories
      • Profiles
      • Startups
      • Tech Titans
      • Partner Content
    • Others
      • Fintech
        • Blockchain
        • Payments
        • Regulation
      • Tech Trends
        • AI
        • AR/VR
        • IoT
      • Video / Podcast
    Blue Tech Wave Media
    Home » China’s Baichuan Intelligence Unveils Open Source Language Model to Compete with OpenAI
    China's-Baichuan-Intelligence-Unveils
    AI

    China’s Baichuan Intelligence Unveils Open Source Language Model to Compete with OpenAI

    By Bal MarsiusJuly 13, 2023Updated:October 3, 2023No Comments3 Mins Read
    Share
    Facebook Twitter LinkedIn Pinterest Email

    Baichuan Intelligence, a startup founded by Wang Xiaochuan, the founder of Sogou, has
    introduced its next-generation large language model Baichuan-13B.
    Baichuan Intelligence, a startup founded by Wang Xiaochuan, the founder of Sogou, has
    introduced its next-generation large language model Baichuan-13B. Wang, a computer science
    prodigy from Tsinghua University, aims to establish China’s version of OpenAI. Baichuan is
    considered one of China’s most promising developers in the field of large language models
    (LLMs). The model, based on the Transformer architecture like OpenAI’s GPT, has 13 billion
    parameters and is trained on Chinese and English data. Baichuan-13B is open source and
    optimised for commercial applications.
    Training Data Comparable to GPT 3.5
    Baichuan-13B is trained on 1.4 trillion tokens, surpassing Meta’s LLaMa, which uses 1 trillion
    tokens in its 13 billion-parameter model. Wang has expressed his intention to release a large-
    scale model comparable to OpenAI’s GPT-3.5 by the end of this year. Within a short period,
    Baichuan has made significant progress, expanding its team to 50 people by the end of April
    and launching its first LLM, Baichuan-7B, in June.​
    Baichuan-13B is now available for free to approved academics and developers who wish to use
    it for commercial purposes. Notably, the model offers variations that can run on consumer-
    grade hardware, addressing the constraints posed by U.S. AI chip sanctions on China.
    Baichuan-7B is an open-source, large-scale pre-training language model, meticulously crafted by
    the visionary minds at Baichuan Intelligent Technology. Rooted in the architecture of the
    Transformer model, this model harnesses a staggering 7 billion parameters and has been
    nourished with the exposure to a staggering 1.2 trillion tokens. With its unwavering versatility,
    Baichuan-7B gracefully accommodates both the Chinese and English languages.
    High Performance Scores Across the Board
    Duly celebrated as a front-runner amongst models of similar scale, Baichuan-7B has emerged
    victorious in renowned Chinese and English benchmarks, including the esteemed C-EVAL and
    MMLU assessments, etching its name at the pinnacle of linguistic excellence.
    This model consistently surpasses its counterparts of similar parameter magnitude, reigning
    supreme as the pre-eminent native pre-trained model in the realm of Chinese language
    comprehension. In the AGIEval assessment, Baichuan-7B outshines other open-source
    contenders, including LLaMA-7B, Falcon-7B, Bloom-7B, and ChatGLM-6B, by an astonishing
    margin, securing an impressive score of 34.4 points.
    Baichuan-7B conquers the C-EVAL examination with a commanding score of 42.8 points,
    outshining ChatGLM-6B’s 38.9 points. In the Gaokao evaluation, the model reigns supreme with
    an exceptional score of 36.2 points, firmly establishing its dominance among pre-trained
    models of comparable parameter scale.
    AGIEval, a celebrated benchmark initiative by Microsoft Research Institute, represents an
    exhaustive endeavour to assess the cognitive and problem-solving capacities of fundamental
    models. C-Eval, a collaborative creation by Shanghai Jiao Tong University, Tsinghua University,
    and the University of Edinburgh, serves as a comprehensive examination evaluating the prowess
    of Chinese language models, encompassing 52 diverse subjects across various industries.
    The Gaokao benchmark, crafted by the esteemed research team at Fudan University, leverages
    the Chinese college entrance examination questions as a dataset, offering a rigorous
    examination of large models’ aptitude in Chinese language comprehension and logical
    reasoning.
    Baichuan-7B’s mastery extends effortlessly into the realm of English. In the esteemed MMLU
    assessment, baichuan-7B astounds with an extraordinary score of 42.5 points, effortlessly
    surpassing the English open-source pre-trained model, LLaMA-7B, and the Chinese open-source
    model, ChatGLM-6B, by significant margins.
    A key determinant of success in large-scale model training resides within the training corpus
    itself. Baichuan Intelligent Technology diligently constructs a high-quality pre-training corpus,
    drawing from rich Chinese learning data and seamlessly integrating high-quality English data.
    This data amalgamation encompasses a vast array of Chinese and English internet data, open-
    source Chinese and English data, alongside a substantial corpus of meticulously curated
    knowledge.

    AI
    Bal Marsius

    Bal was BTW's copywriter specialising in tech and productivity tools. He has experience working in startups, mid-size tech companies, and non-profits.

    Related Posts

    Interview with Sarath Babu Rayaprolu from Voxtera on dynamic and secure VoIP

    July 7, 2025

    Authors sue Microsoft over AI training using their books

    July 3, 2025

    OpenAI clarifies Robinhood tokens are not real equity

    July 3, 2025
    Add A Comment
    Leave A Reply Cancel Reply

    CATEGORIES
    Archives
    • July 2025
    • June 2025
    • May 2025
    • April 2025
    • March 2025
    • February 2025
    • January 2025
    • December 2024
    • November 2024
    • October 2024
    • September 2024
    • August 2024
    • July 2024
    • June 2024
    • May 2024
    • April 2024
    • March 2024
    • February 2024
    • January 2024
    • December 2023
    • November 2023
    • October 2023
    • September 2023
    • August 2023
    • July 2023

    Blue Tech Wave (BTW.Media) is a future-facing tech media brand delivering sharp insights, trendspotting, and bold storytelling across digital, social, and video. We translate complexity into clarity—so you’re always ahead of the curve.

    BTW
    • About BTW
    • Contact Us
    • Join Our Team
    TERMS
    • Privacy Policy
    • Cookie Policy
    • Terms of Use
    Facebook X (Twitter) Instagram YouTube LinkedIn

    Type above and press Enter to search. Press Esc to cancel.