国产98在线 | 传媒麻豆,久久―日本道色综合久久,久久久精品久久久久久久久久久,国产成人精品免费久久久久,五月综合色婷婷影院在线观看,久久久亚洲精品视频

  • 虎科技 - 領先的互聯(lián)網(wǎng)科技媒體

硅基流動發(fā)布DeepSeek-V3.1,支持160K超長上下文,性能大幅提升

   時間:2025-08-26 03:04 來源:ITBEAR作者:任飛揚

硅基流動公司近日正式宣布,其平臺上已部署了深度求索團隊最新研發(fā)的開源模型——DeepSeek-V3.1。這一消息標志著人工智能領域的一次重要進展。

DeepSeek-V3.1在參數(shù)規(guī)模上實現(xiàn)了突破,總參數(shù)達到6710億,其中激活參數(shù)為370億。該模型采用了創(chuàng)新的混合推理架構,能夠靈活切換思考模式與非思考模式,相較于之前的版本DeepSeek-R1-0528,在思考效率上有了顯著提升,能夠在更短的時間內給出高質量的答案。DeepSeek-V3.1在工具使用和智能體任務中的表現(xiàn)也取得了長足的進步。

在性能測試方面,DeepSeek-V3.1在SWE、Browsecomp、xbench和SimpleQA等基礎測試上均實現(xiàn)了大幅度的性能提升,其表現(xiàn)已經(jīng)與Claude 4 Sonnet等頂尖模型相媲美。這一成就不僅展示了DeepSeek-V3.1的強大實力,也進一步鞏固了硅基流動在人工智能領域的領先地位。

在價格方面,DeepSeek-V3.1也展現(xiàn)出了極高的性價比。其輸入價格為每百萬標記(M Tokens)4元,輸出價格為每百萬標記12元,相較于Claude 4 Sonnet,輸入成本便宜了5倍,輸出成本更是便宜了9倍。這一價格優(yōu)勢無疑將吸引更多的用戶選擇DeepSeek-V3.1。

DeepSeek-V3.1還支持了前所未有的160K超長上下文處理,這一特性使其在處理長文檔、多輪對話、編碼及智能體等復雜場景時具有得天獨厚的優(yōu)勢。例如,在處理超長文檔時,用戶無需再進行繁瑣的分段切割,DeepSeek-V3.1能夠無損解析法律合同比對、學術文獻綜述、金融報告解析以及全量代碼庫理解等任務。同時,該模型還具備深度對話長程記憶的能力,能夠適應復雜客服、心理咨詢、教學輔導等強連貫性交互場景。DeepSeek-V3.1還能夠實現(xiàn)復雜任務的端到端處理,告別分段處理或緩存拼接等繁瑣的工程邏輯,一站式完成全鏈路任務。

在檢索增強方面,DeepSeek-V3.1也展現(xiàn)出了卓越的性能。它能夠將百份文檔嵌入上下文,并基于完整信息生成精準答案,從而顯著降低檢索錯誤率。這一特性使得DeepSeek-V3.1在信息檢索和問答系統(tǒng)等領域具有廣泛的應用前景。

 
 
更多>同類內容
推薦圖文
推薦內容
點擊排行
 
智快科技微信賬號
ITBear微信賬號

微信掃一掃
加微信拉群
電動汽車群
科技數(shù)碼群

主站蜘蛛池模板: | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |