{"id":1161,"date":"2025-08-08T18:46:47","date_gmt":"2025-08-08T10:46:47","guid":{"rendered":"https:\/\/www.ygcloud.com\/knowledge\/?p=1161"},"modified":"2025-08-08T18:46:47","modified_gmt":"2025-08-08T10:46:47","slug":"gpt-5%e9%a6%96%e5%8f%91%e8%af%84%e6%b5%8b%ef%bc%9a%e6%8e%a8%e7%90%86%e9%94%99%e8%af%af%e7%8e%87%e6%9a%b4%e9%99%8d80%ef%bc%8c%e5%ae%9e%e6%93%8d%e6%8c%87%e5%8d%97%e8%ae%b2%e8%a7%a3","status":"publish","type":"post","link":"https:\/\/www.ygcloud.com\/knowledge\/1161\/","title":{"rendered":"GPT-5\u9996\u53d1\u8bc4\u6d4b\uff1a\u63a8\u7406\u9519\u8bef\u7387\u66b4\u964d80%\uff0c\u5b9e\u64cd\u6307\u5357\u8bb2\u89e3"},"content":{"rendered":"

<\/p>\n

AI\u9769\u547d\u8fce\u6765\u5386\u53f2\u6027\u65f6\u523b\uff01<\/span> \u4eca\u65e5\u51cc\u6668\uff0cOpenAI \u6b63\u5f0f\u53d1\u5e03 GPT-5\uff0c\u5168\u7403\u5f00\u53d1\u8005\u793e\u533a\u77ac\u95f4\u6cb8\u817e\u3002\u8fd9\u6b21\u5347\u7ea7\u4e0d\u4ec5\u662f\u6a21\u578b\u80fd\u529b\u7684\u98de\u8dc3\uff0c\u66f4\u98a0\u8986\u4e86\u4eba\u673a\u4ea4\u4e92\u903b\u8f91\u2014\u2014GPT-5 \u9996\u6b21\u8ba9AI\u62e5\u6709\u201c\u81ea\u4e3b\u601d\u8003\u201d\u80fd\u529b<\/span>\uff0c\u63a8\u7406\u9519\u8bef\u7387\u9aa4\u964d80%\uff0c\u590d\u6742\u4efb\u52a1\u5904\u7406\u6548\u7387\u7ffb\u500d\u3002\u7b14\u8005\u7b2c\u4e00\u65f6\u95f4\u5b9e\u6d4b\u5e76\u89e3\u6790\u6280\u672f\u6587\u6863\uff0c\u5e26\u4f60\u76f4\u51fb\u8fd9\u573aAI\u8fdb\u5316\u98ce\u66b4\u7684\u6838\u5fc3\uff01<\/p>\n

\"\"<\/p>\n

\u4e00\u3001\u98a0\u8986\u6027\u67b6\u6784\uff1aAI\u7684\u201c\u53cc\u8111\u7cfb\u7edf\u201d<\/span><\/h3>\n

GPT-5 \u4e0d\u518d\u662f\u5355\u4e00\u6a21\u578b\uff0c\u800c\u662f\u7531\u4e09\u5927\u6a21\u5757\u7ec4\u6210\u7684\u667a\u80fd\u51b3\u7b56\u4e2d\u67a2<\/span>\uff1a<\/p>\n\n\n\n\n\n\n\n
\u6a21\u5757<\/th>\n\u529f\u80fd\u8bf4\u660e<\/th>\n<\/tr>\n<\/thead>\n
\ud83d\ude80 \u9ad8\u6548\u5e94\u7b54\u6a21\u578b<\/span><\/td>\n\u5b9e\u65f6\u54cd\u5e94\u65e5\u5e38\u5bf9\u8bdd\u3001\u5199\u4f5c\u7b49\u8f7b\u91cf\u4efb\u52a1\uff08\u901f\u5ea6\u6bd4GPT-4o\u5feb40%\uff09<\/td>\n<\/tr>\n
\ud83e\udde0 Thinking\u6df1\u5ea6\u6a21\u5f0f<\/span><\/td>\n\u4e13\u653b\u6570\u5b66\u8bc1\u660e\u3001\u4ee3\u7801\u8c03\u8bd5\u7b49\u590d\u6742\u95ee\u9898\uff0c\u81ea\u52a8\u6784\u5efa\u601d\u7ef4\u94fe\u964d\u4f4e\u9519\u8bef\u7387<\/td>\n<\/tr>\n
\ud83d\udd00 \u667a\u80fd\u8def\u7531Router<\/span><\/td>\n\u6839\u636e\u95ee\u9898\u96be\u5ea6\u81ea\u52a8\u5207\u6362\u6a21\u578b\uff0c\u5b9e\u73b0\u201c\u5feb\u95ee\u5feb\u7b54\uff0c\u96be\u95ee\u6df1\u60f3\u201d\u7684\u667a\u80fd\u8c03\u5ea6<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

\u5b9e\u6d4b\u6848\u4f8b\uff1a<\/strong>
\n\u5f53\u8981\u6c42\u300c\u89e3\u6790\u4e00\u7bc7\u91cf\u5b50\u8ba1\u7b97\u8bba\u6587\u7684\u6838\u5fc3\u516c\u5f0f\u300d\u65f6\uff0cRouter \u81ea\u52a8\u6fc0\u6d3b Thinking \u6a21\u5f0f\uff1a
\n1\ufe0f\u20e3 \u5148\u62c6\u89e3\u516c\u5f0f\u7b26\u53f7\u5b9a\u4e49
\n2\ufe0f\u20e3 \u9010\u6b65\u63a8\u5bfc\u6570\u5b66\u903b\u8f91\u94fe
\n3\ufe0f\u20e3 \u6700\u540e\u7528\u767d\u8bdd\u89e3\u91ca\u7269\u7406\u610f\u4e49
\n\u5bf9\u6bd4GPT-4o\u7684\u8df3\u6b65\u56de\u7b54\uff0c\u9519\u8bef\u7387\u4ece19.3%\u964d\u81f33.7%<\/span>\uff08\u6765\u6e90\uff1aGPQA\u79d1\u7814\u57fa\u51c6\u6d4b\u8bd5\uff09<\/p>\n

\u4e8c\u3001\u6027\u80fd\u66b4\u589e\uff1a80%\u9519\u8bef\u7387\u4e0b\u964d\u5982\u4f55\u5b9e\u73b0\uff1f<\/span><\/h3>\n

\u2705 \u63a8\u7406\u80fd\u529b\u53f2\u8bd7\u7ea7\u7a81\u7834<\/span><\/h4>\n
    \n
  • AIME 2025\u6570\u5b66\u7ade\u8d5b<\/span>\uff1a94.6%\u6b63\u786e\u7387\uff08\u4eba\u7c7b\u91d1\u724c\u9009\u624b\u5e73\u5747\u5206\u4ec585%\uff09<\/li>\n
  • \u7f16\u7a0b\u5b9e\u6218\u6d4b\u8bd5<\/span>\uff1a\u5728SWE-bench\u4e2d\u4fee\u590d\u4ee3\u7801\u9519\u8bef\u6210\u529f\u738774.9%<\/span>\uff08GPT-4o\u4ec530.8%\uff09<\/li>\n
  • \u533b\u5b66\u8bca\u65ad<\/span>\uff1aHealthBench Hard\u6d4b\u8bd5\u5e7b\u89c9\u73871.6%<\/span>\uff08\u8f83\u524d\u4ee3\u964d80%\uff09\uff0c\u53ef\u89e3\u6790CT\u62a5\u544a\u4f46\u660e\u793a\u201c\u975e\u533b\u7597\u8bca\u65ad\u201d<\/li>\n<\/ul>\n

    \ud83c\udf1f Thinking\u6a21\u5f0f\u8fd0\u4f5c\u63ed\u79d8<\/span><\/h4>\n

    \u5f53\u7528\u6237\u63d0\u95ee\u590d\u6742\u95ee\u9898\u65f6\uff1a<\/p>\n

    graph LR A[\u7528\u6237\u8f93\u5165\u95ee\u9898] –> B{Router\u667a\u80fd\u5224\u65ad} B –>|\u7b80\u5355\u95ee\u9898| C[\u9ad8\u6548\u5e94\u7b54\u6a21\u578b-\u79d2\u7ea7\u54cd\u5e94] B –>|\u590d\u6742\u95ee\u9898| D[Thinking\u6a21\u5f0f-\u542f\u52a8\u591a\u6b65\u63a8\u7406] D –> E[\u62c6\u89e3\u95ee\u9898\u5b50\u4efb\u52a1] E –> F[\u8c03\u7528\u5de5\u5177\u94fe\u9a8c\u8bc1] F –> G[\u751f\u6210\u4eba\u7c7b\u53ef\u9a8c\u8bc1\u7684\u601d\u8003\u8fc7\u7a0b]<\/div>\n

    \u7ed3\u679c<\/span>\uff1a\u903b\u8f91\u6f0f\u6d1e\u51cf\u5c1180%\uff0c\u7b54\u6848\u53ef\u8ffd\u6eaf\u6027\u63d0\u5347300%\uff08OpenAI\u6280\u672f\u767d\u76ae\u4e66\uff09<\/p>\n

    \u4e09\u3001\u514d\u8d39\u4f7f\u7528\u6307\u5357\uff1a\u624b\u628a\u624b\u6559\u4f60\u73a9\u8f6cGPT-5<\/span><\/h3>\n

    \ud83d\udd25 \u96f6\u6210\u672c\u4f53\u9a8c\u8def\u5f84<\/span><\/h4>\n\n\n\n\n\n\n\n
    \u65b9\u5f0f<\/th>\n\u64cd\u4f5c\u6b65\u9aa4<\/th>\n\u529f\u80fd\u6743\u9650<\/th>\n<\/tr>\n<\/thead>\n
    \u7f51\u9875\u514d\u8d39\u7248<\/span><\/td>\n\u767b\u5f55 ChatGPT\u5b98\u7f51<\/a><\/td>\n\u9ed8\u8ba4\u4f7f\u7528GPT-5\u7cbe\u7b80\u7248\uff08\u5f3a\u4e8eGPT-4o\uff09<\/td>\n<\/tr>\n
    API\u767d\u5ad6<\/span><\/td>\n\u65b0\u6ce8\u518c\u5f00\u53d1\u8005\u83b7$5\u8bd5\u7528\u91d1 \u2192 \u8c03\u7528gpt-5-mini<\/span>\u200b\u6a21\u578b<\/td>\n\u9002\u5408\u8f7b\u91cf\u5f00\u53d1\u6d4b\u8bd5<\/td>\n<\/tr>\n
    Azure\u5165\u53e3<\/span><\/td>\n\u7533\u8bf7 Azure OpenAI<\/a> \u2192 \u7ed1\u5b9a\u4f01\u4e1a\u8d26\u53f7\u9886\u53d6\u989d\u5ea6<\/td>\n\u4f01\u4e1a\u7ea7\u7a33\u5b9a\u63a5\u5165<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    \ud83d\udca1 \u9ad8\u9636\u7528\u6237\u6280\u5de7<\/h4>\n
    # \u901a\u8fc7API\u5f3a\u5236\u542f\u7528Thinking\u6a21\u5f0f\uff08\u4ec5Pro\u7248\u672c\u652f\u6301\uff09  \r\nimport openai  \r\nresponse = openai.ChatCompletion.create(  \r\n  model=\"gpt-5-pro\",  \r\n  messages=[{\"role\": \"user\", \"content\": \"\u8bc1\u660e\u9ece\u66fc\u731c\u60f3\u7b2c\u4e09\u6b65\u7684\u53ef\u884c\u6027\uff1f\"}],  \r\n  thinking_mode=True  # \u6fc0\u6d3b\u6df1\u5ea6\u63a8\u7406  \r\n)  \r\n<\/code><\/pre>\n
    \u56db\u3001\u6a2a\u5411\u8bc4\u6d4b\uff1aGPT-5 vs \u4e3b\u6d41\u6a21\u578b<\/span><\/h3>\n\n\n\n\n\n\n\n\n
    \u80fd\u529b\u9879<\/th>\nGPT-5<\/th>\nGPT-4o<\/th>\nClaude 3 Opus<\/th>\n<\/tr>\n<\/thead>\n
    \u590d\u6742\u63a8\u7406<\/span><\/td>\n\ud83d\udfe2 Thinking\u6a21\u5f0f\u5236\u80dc<\/td>\n\ud83d\udfe1 \u6613\u8df3\u6b65<\/td>\n\ud83d\udfe2 \u7a33\u5b9a\u4f46\u7f13\u6162<\/td>\n<\/tr>\n
    \u591a\u6a21\u6001\u7406\u89e3<\/span><\/td>\n\u2705 \u56fe\u50cf\/\u89c6\u9891\/\u8bed\u97f3<\/td>\n\u2705 \u56fe\u50cf\/\u8bed\u97f3<\/td>\n\u274c \u4ec5\u6587\u672c<\/td>\n<\/tr>\n
    \u7f16\u7a0b\u5b9e\u6218<\/span><\/td>\n74.9%(SWE-bench)<\/td>\n30.8%<\/td>\n68.2%<\/td>\n<\/tr>\n
    \u6210\u672c\u6548\u76ca<\/span><\/td>\n$10\/\u767e\u4e07\u8f93\u51fatoken<\/td>\n$5<\/td>\n$15<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n
    \n
    \u7b14\u8005\u76f4\u8a00<\/span>\uff1aGPT-5\u5728\u79d1\u7814\u3001\u7f16\u7a0b\u7b49\u4e13\u4e1a\u573a\u666f\u4f18\u52bf\u663e\u8457\uff0c\u4f46\u65e5\u5e38\u4f7f\u7528\u514d\u8d39\u7248\u5df2\u8db3\u591f\u78be\u538b\u524d\u4ee3\uff01<\/p>\n<\/blockquote>\n
    \u4e94\u3001\u9769\u547d\u6027\u5f71\u54cd\uff1aAI\u4ece\u6b64\u201c\u50cf\u4eba\u7c7b\u4e00\u6837\u601d\u8003\u201d<\/span><\/h3>\n
    \u5f53GPT-5\u5728\u53d1\u5e03\u4f1a\u73b0\u573a\u5b8c\u6210\u4ee5\u4e0b\u4efb\u52a1\u65f6\uff0c\u89c2\u4f17\u96c6\u4f53\u60ca\u547c\uff1a
    \n1\ufe0f\u20e3 \u533b\u5b66\u8f85\u52a9<\/span>\uff1a\u89e3\u6790\u7cd6\u5c3f\u75c5\u60a3\u8005\u8840\u7cd6\u6ce2\u52a8\u56fe\u8c31\uff0c\u751f\u6210\u996e\u98df\u8c03\u6574\u65b9\u6848
    \n2\ufe0f\u20e3 \u8de8\u5b66\u79d1\u7814\u7a76<\/span>\uff1a\u7528\u62d3\u6251\u5b66\u539f\u7406\u4f18\u5316\u57ce\u5e02\u5730\u94c1\u89c4\u5212\uff08\u9644\u53d7\u529b\u6a21\u62df\u4ee3\u7801\uff09
    \n3\ufe0f\u20e3 \u521b\u610f\u7a81\u7834<\/span>\uff1a\u57fa\u4e8e\u68b5\u9ad8\u753b\u98ce\u751f\u6210\u4ea4\u4e92\u5f0f\u5149\u5f71\u88c5\u7f6e\u8bbe\u8ba1\u7a3f<\/p>\n
    Altman\u73b0\u573a\u65ad\u8a00<\/span>\uff1a<\/p>\n
    \n
    \u201cGPT-5\u7684Thinking\u6a21\u5f0f\u4e0d\u662f\u7b54\u6848\u673a\uff0c\u800c\u662f\u63a8\u7406\u4f19\u4f34<\/span>\u2014\u2014\u5b83\u6b63\u5728\u903c\u8fd1\u4eba\u7c7b\u4e13\u5bb6\u7ea7\u7684\u601d\u7ef4\u94fe\u6761\u3002\u201d<\/p>\n<\/blockquote>\n
    \u7acb\u5373\u884c\u52a8\u6307\u5357<\/span><\/h3>\n
    \ud83c\udf1f **\u63a8\u8350\u4eba\u7fa4**\uff1a  \r\n- \u7814\u7a76\u8005\/\u5de5\u7a0b\u5e08 \u2192 \u5fc5\u7528Pro\u7248API\u89e3\u9501Thinking\u6a21\u5f0f  \r\n- \u5b66\u751f\/\u521b\u4f5c\u8005 \u2192 \u514d\u8d39\u7248\u7f51\u9875\u7aef\u4f53\u9a8c\u667a\u80fd\u98de\u8dc3  \r\n- \u4f01\u4e1a\u5ba2\u6237 \u2192 \u901a\u8fc7Azure\u90e8\u7f72\u5b89\u5168\u79c1\u6709\u5316\u7248\u672c  \r\n\r\n\ud83d\ude80 **\u76f4\u8fbe\u5165\u53e3**\uff1a  \r\n- [ChatGPT\u5b98\u7f51](https:\/\/chat.openai.com)  \r\n- [API\u5e73\u53f0](https:\/\/platform.openai.com)  \r\n<\/code><\/pre>\n
    \u5212\u91cd\u70b9<\/span>\uff1a\u514d\u8d39\u7528\u6237\u5373\u65e5\u5373\u53ef\u4f53\u9a8c\u5347\u7ea7\u7248\u63a8\u7406\u80fd\u529b\uff0c\u65e0\u9700\u7b49\u5f85\uff01<\/p>\n
    \n
    \u8fd9\u573a\u201c\u601d\u8003\u9769\u547d\u201d\u624d\u521a\u521a\u5f00\u59cb\u2014\u2014\u4f60\u7684AI\u52a9\u624b\uff0c\u7ec8\u4e8e\u5b66\u4f1a\u201c\u4e09\u601d\u800c\u540e\u884c\u201d\u4e86\u3002<\/p>\n<\/blockquote>\n","protected":false},"excerpt":{"rendered":"
    AI\u9769\u547d\u8fce\u6765\u5386\u53f2\u6027\u65f6\u523b\uff01 \u4eca\u65e5\u51cc\u6668\uff0cOpenAI \u6b63\u5f0f\u53d1\u5e03 GPT-5\uff0c\u5168\u7403\u5f00\u53d1\u8005\u793e\u533a\u77ac\u95f4\u6cb8\u817e\u3002\u8fd9\u6b21\u5347\u7ea7\u4e0d\u4ec5\u662f\u6a21\u578b\u80fd\u529b\u7684\u98de\u8dc3\uff0c\u66f4\u98a0\u8986\u4e86\u4eba\u673a\u4ea4\u4e92\u903b\u8f91\u2014\u2014GPT-5 \u9996\u6b21\u8ba9AI\u62e5\u6709\u201c\u81ea\u4e3b\u601d\u8003\u201d\u80fd\u529b\uff0c\u63a8\u7406\u9519\u8bef\u7387\u9aa4\u964d80%\uff0c\u590d\u6742\u4efb\u52a1\u5904\u7406\u6548\u7387\u7ffb\u500d\u3002\u7b14…<\/p>\n","protected":false},"author":1,"featured_media":1162,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[95],"tags":[97],"topic":[],"class_list":["post-1161","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ai","tag-ai"],"_links":{"self":[{"href":"https:\/\/www.ygcloud.com\/knowledge\/wp-json\/wp\/v2\/posts\/1161","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.ygcloud.com\/knowledge\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.ygcloud.com\/knowledge\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.ygcloud.com\/knowledge\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.ygcloud.com\/knowledge\/wp-json\/wp\/v2\/comments?post=1161"}],"version-history":[{"count":1,"href":"https:\/\/www.ygcloud.com\/knowledge\/wp-json\/wp\/v2\/posts\/1161\/revisions"}],"predecessor-version":[{"id":1163,"href":"https:\/\/www.ygcloud.com\/knowledge\/wp-json\/wp\/v2\/posts\/1161\/revisions\/1163"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.ygcloud.com\/knowledge\/wp-json\/wp\/v2\/media\/1162"}],"wp:attachment":[{"href":"https:\/\/www.ygcloud.com\/knowledge\/wp-json\/wp\/v2\/media?parent=1161"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.ygcloud.com\/knowledge\/wp-json\/wp\/v2\/categories?post=1161"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.ygcloud.com\/knowledge\/wp-json\/wp\/v2\/tags?post=1161"},{"taxonomy":"topic","embeddable":true,"href":"https:\/\/www.ygcloud.com\/knowledge\/wp-json\/wp\/v2\/topic?post=1161"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}