|
中国のテック大手Tencentは、2026年8月28日に新世代のフラッグシップ大規模言語モデル「Hy4 preview」(テンセント混元/Hunyuan Hy4)を発表し、オープンソース(オープンウェイト)として一般公開しました。 Hy4 previewはMixture-of-Experts、略してMoEという方式で、すべての計算部分を毎回動かすのではなく、入力に応じて必要な専門部分だけを動かします。総パラメータ数は7700億ですが、1トークンの処理で稼働するのは490億。 このHy4-previewについて生成AIに深掘りさせましたのでご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。 Tencent Announces “Hy4 Preview” Chinese technology giant Tencent announced its next-generation flagship large language model, “Hy4 Preview”—also known as Tencent Hunyuan Hy4—on August 28, 2026, and made it publicly available as an open-source, or more precisely open-weight, model. Hy4 Preview employs a Mixture-of-Experts (MoE) architecture. Rather than activating the entire model for every computation, it selectively activates only the expert components needed for each input. Although the model has 770 billion parameters in total, only 49 billion parameters are activated when processing each token. I asked generative AI to conduct an in-depth analysis of Hy4 Preview, and the results are provided for your reference. Please note, however, that the research and analysis generated by AI are based solely on publicly available information, may not necessarily reflect the actual situation, and may contain inaccuracies. Your browser does not support viewing this document. Click here to download the document. Your browser does not support viewing this document. Click here to download the document. Your browser does not support viewing this document. Click here to download the document. Your browser does not support viewing this document. Click here to download the document. Your browser does not support viewing this document. Click here to download the document. Your browser does not support viewing this document. Click here to download the document. Your browser does not support viewing this document. Click here to download the document. Your browser does not support viewing this document. Click here to download the document. Your browser does not support viewing this document. Click here to download the document.
0 Comments
Leave a Reply. |
著者萬秀憲 アーカイブ
April 2026
カテゴリー |
RSS Feed