Googleは12月19日、新たな推論AIモデル「Gemini 2.0 Flash Thinking」を発表し、Experimental(実験)モデルをGoogle AI Studioで公開しました。12月11日に登場したばかりのGemini 2.0 Flashの派生モデルです。
Gemini 2.0 Flash Thinkingは、回答時に「思考プロセス(thinking process)」を生成、ベースとした「Gemini 2.0 Flash」よりも高度な推論が可能ということで、OpenAI o1のライバルと言えそうです。 各種テストでOpenAIのo1-previewやGPT-4oを超える性能という評価や、Gemini 2.0 Flash ThinkingとOpenAI o1を比較してみた結果、OpenAI o1の圧勝だったという報告もあります。 性能競争が新たな段階に入ったようです。 Googleが思考プロセスを導入して推論を強化したAIモデル「Gemini 2.0 Flash Thinking」を公開、各種テストでOpenAIのo1-previewやGPT-4oを超える性能 https://gigazine.net/news/20241220-google-gemini-flash-thinking/ GoogleのGemini 2.0 Flash Thinkingを徹底検証!思考プロセスが生む推論力とは? 2024-12-20 https://weel.co.jp/media/tech/gemini-2-flash-thinking/ Google、「Gemini 2.0 Flash Thinking」公開 「o1」競合の“思考過程を示す”推論AIモデル https://www.itmedia.co.jp/news/articles/2412/20/news108.html New Inference AI Model "Gemini 2.0 Flash Thinking" On December 19, Google announced a new inference AI model called "Gemini 2.0 Flash Thinking" and released an experimental version in Google AI Studio. This model is a derivative of "Gemini 2.0 Flash," which was introduced just recently on December 11. Gemini 2.0 Flash Thinking is said to enable more advanced reasoning compared to its predecessor, "Gemini 2.0 Flash," by generating a "thinking process" during responses. It is expected to rival OpenAI's o1 model. Performance evaluations indicate that it has surpassed OpenAI's o1-preview and GPT-4o in various tests. However, there are also reports that OpenAI's o1 achieved overwhelming victories when directly compared with Gemini 2.0 Flash Thinking. It seems the competition in AI performance has entered a new phase.
0 Comments
Leave a Reply. |
著者萬秀憲 アーカイブ
February 2025
カテゴリー |