{"id":790039,"date":"2026-05-31T10:11:38","date_gmt":"2026-05-31T02:11:38","guid":{"rendered":"https:\/\/www.alphaseek.cn\/?p=790039"},"modified":"2026-05-31T10:11:38","modified_gmt":"2026-05-31T02:11:38","slug":"openai%e5%ae%98%e5%ae%a3%e9%80%80%e5%bd%b9o3%e4%b8%8egpt-4-5","status":"publish","type":"post","link":"https:\/\/www.alphaseek.cn\/?p=790039","title":{"rendered":"OpenAI\u5b98\u5ba3\u9000\u5f79o3\u4e0eGPT-4.5"},"content":{"rendered":"

\"OpenAI\u5b98\u5ba3\u9000\u5f79o3\u4e0eGPT-4.5\"<\/p>\n

\"OpenAI\u5b98\u5ba3\u9000\u5f79o3\u4e0eGPT-4.5\"<\/p>\n

o3\u88ab\u5c01\u300cGOAT\u300d\u3001GPT-4.5\u88ab\u53eb\u300c\u7075\u9b42\u5199\u624b\u300d\uff0cOpenAI<\/a>\u8bf4\u9000\u5c31\u9000\u3002GPT-5.6\u5df2\u5728\u70ed\u8eab\u2014\u2014\u4f46\u300c\u66f4\u5f3a\u300d\u80fd\u4e0d\u80fd\u4fe1\uff1fOpenAI\u81ea\u5df1\u8bf4\uff1a\u672a\u5fc5\u3002<\/strong><\/p>\n

OpenAI\u53c8\u5e72\u4e86\u4ef6\u8ba9\u8001\u7528\u6237\u5fc3\u788e\u7684\u4e8b\u3002<\/p>\n

28\u53f7\uff0c\u4ed6\u4eec\u5728\u5b98\u65b9Release Notes\u4e0a\u5ba3\u5e03\uff1a\u4ece2026\u5e748\u670826\u65e5\u8d77\uff0co3\u4eceChatGPT<\/a>\u6b63\u5f0f\u9000\u5f79\uff01<\/strong><\/p>\n

GPT-4.5\u66f4\u72e0\uff0c6\u670827\u65e5\u5c31\u4e0b\u7ebf<\/strong>\uff0c\u53ea\u7ed930\u5929\u7f13\u51b2\u3002<\/p>\n

\"OpenAI\u5b98\u5ba3\u9000\u5f79o3\u4e0eGPT-4.5\"<\/p>\n

\u4e24\u4e2a\u6a21\u578b\u76ee\u524d\u4ec5\u9650\u4ed8\u8d39\u7528\u6237\u5728\u8bbe\u7f6e\u91cc\u624b\u52a8\u5207\u6362\u624d\u80fd\u7528\u5230\u2014\u2014\u5b83\u4eec\u65e9\u5c31\u4e0d\u662f\u9ed8\u8ba4\u9009\u9879\u4e86\uff0c\u8fd9\u6b21\u662f\u8981\u5f7b\u5e95\u4ece\u83dc\u5355\u4e0a\u62b9\u6389\u3002<\/p>\n

\"OpenAI\u5b98\u5ba3\u9000\u5f79o3\u4e0eGPT-4.5\"<\/p>\n

\u4e0e\u6b64\u540c\u65f6\uff0c\u77e5\u540d\u535a\u4e3bLeo 5\u670829\u65e5\u53d1\u5e16\u786e\u8ba4\uff1aGPT-5.6\u7684\u5f00\u53d1\u5168\u529b\u63a8\u8fdb\u4e2d\uff0c\u4e00\u4e2a\u660e\u663e\u66f4\u5f3a\u7684\u65b0\u68c0\u67e5\u70b9\u5df2\u5728\u5185\u90e8\u4e0a\u7ebf\u3002<\/p>\n

\"OpenAI\u5b98\u5ba3\u9000\u5f79o3\u4e0eGPT-4.5\"<\/p>\n

\u4e00\u8fb9\u662f\u7528\u6237\u8fd8\u6ca1\u6765\u5f97\u53ca\u8bf4\u518d\u89c1\uff0c\u4e00\u8fb9\u662f\u65b0\u6a21\u578b\u5df2\u7ecf\u5728\u540e\u53f0\u70ed\u8eab\u3002<\/p>\n

\u6700\u540e\u4e24\u4e2a\u597d\u7528\u7684\u6a21\u578b\uff0c\u8981\u6ca1\u4e86<\/strong><\/p>\n

\u8fd9\u8bdd\u4e0d\u662f\u6211\u8bf4\u7684\u3002<\/p>\n

\"OpenAI\u5b98\u5ba3\u9000\u5f79o3\u4e0eGPT-4.5\"<\/p>\n

Reddit\u70ed\u5e16\u4e0b\u9762\u51e0\u767e\u6761\u8bc4\u8bba\uff0c\u4ed8\u8d39\u7528\u6237\u4eec\u5728ChatGPT\u8bbe\u7f6e\u91cc\u75af\u72c2\u622a\u56fe\u7559\u5ff5\uff0c\u6c1b\u56f4\u50cf\u662f\u8d76\u5728\u62c6\u8fc1\u524d\u6700\u540e\u4e00\u6b21\u62cd\u7167\u3002<\/p>\n

\"OpenAI\u5b98\u5ba3\u9000\u5f79o3\u4e0eGPT-4.5\"<\/p>\n

\u6709\u7f51\u53cb\u6124\u6012\u5730\u8868\u793a\uff0cOpenAI\u8fd9\u79cd\u64cd\u4f5c\u300c\u7ed9\u4e86\u6211\u53d6\u6d88Pro\u8d26\u6237\u7684\u6700\u540e\u4e00\u4e2a\u7406\u7531\u3002\u300d<\/p>\n

\"OpenAI\u5b98\u5ba3\u9000\u5f79o3\u4e0eGPT-4.5\"<\/p>\n

\u8fd8\u6709\u7528\u6237\u8868\u8fbe\u4e86\u5bf9o3\u5199\u4f5c\u98ce\u683c\u7684\u7559\u604b\u4e0e\u4e0d\u820d\u3002<\/p>\n

\"OpenAI\u5b98\u5ba3\u9000\u5f79o3\u4e0eGPT-4.5\"<\/p>\n

o3\u662fOpenAI\u7684\u300c\u7eaf\u8840\u63a8\u7406\u6a21\u578b\u300d\uff0c2025\u5e744\u6708\u4e0a\u7ebf\uff0c\u4e13\u95e8\u7ed9\u90a3\u4e9b\u9700\u8981\u4e00\u6b65\u6b65\u60f3\u3001\u4e00\u5c42\u5c42\u5265\u7684\u786c\u6838\u4efb\u52a1\u8bbe\u8ba1\u7684\u3002<\/p>\n

\u6570\u5b66\u8bc1\u660e\u3001\u79d1\u5b66\u63a8\u5bfc\u3001\u4ee3\u7801debug\u2026\u2026\u5728\u8fd9\u4e9b\u9700\u8981\u300c\u6162\u601d\u8003\u300d\u7684\u573a\u666f\u91cc\uff0co3\u51e0\u4e4e\u662f\u65e0\u654c\u7684\u5b58\u5728\u3002<\/p>\n

\"OpenAI\u5b98\u5ba3\u9000\u5f79o3\u4e0eGPT-4.5\"<\/p>\n

X\u4e0a\u6709\u7528\u6237\u76f4\u63a5\u5c01\u5b83\u4e3a\u300cGOAT\u300d\uff08Greatest Of All Time\uff0c\u53f2\u4e0a\u6700\u5f3a\uff09\u3002<\/p>\n

\u8fd8\u6709\u4eba\u8bf4\u5f97\u66f4\u76f4\u767d\uff1a\u300co3\u662f\u6700\u540e\u4e00\u4e2a\u771f\u6b63\u5728\u300e\u60f3\u95ee\u9898\u300f\u7684\u6a21\u578b\uff0c5\u7cfb\u5217\u66f4\u806a\u660e\uff0c\u4f46\u5c11\u4e86\u90a3\u80a1\u5b50\u8f74\u52b2\u513f\u3002\u300d<\/p>\n

\"OpenAI\u5b98\u5ba3\u9000\u5f79o3\u4e0eGPT-4.5\"<\/p>\n

o3-pro\u66f4\u662fPro\u7528\u6237\u7684\u5fc3\u5934\u597d\u2014\u2014\u82b1\u66f4\u591a\u65f6\u95f4\u601d\u8003\uff0c\u7ed9\u51fa\u66f4\u53ef\u9760\u7684\u7b54\u6848\uff0c\u5728\u6570\u5b66\u3001\u79d1\u5b66\u3001\u7f16\u7a0b\u4e09\u5927\u9886\u57df\u78be\u538b\u5f0f\u9886\u5148\uff0c\u5b66\u672f\u8bc4\u4f30\u91cc\u7684\u6210\u7ee9\u5355\u6f02\u4eae\u5f97\u50cf\u5b66\u9738\u7684\u671f\u672b\u8003\u3002<\/p>\n

\u518d\u8bf4GPT-4.5\u3002<\/strong><\/p>\n

\u5982\u679c\u8bf4o3\u662f\u7406\u5de5\u79d1\u5b66\u9738\uff0cGPT-4.5\u5c31\u662f\u6587\u79d1\u5929\u624d\u3002<\/p>\n

\u8fd9\u662f\u8bb8\u591a\u7528\u6237\u5fc3\u4e2d\u300c\u5199\u4f5c\u6700\u81ea\u7136\u300d\u7684\u6a21\u578b\uff0c\u6587\u5b57\u6709\u6e29\u5ea6\u3001\u6709\u8282\u594f\u611f\u3001\u6709\u7075\u9b42\u3002<\/p>\n

X\u7528\u6237Striver\u7684\u8bc4\u4ef7\u88ab\u5e7f\u6cdb\u8f6c\u53d1\uff1a\u76f4\u5230\u4eca\u5929\uff0c4.5\u4ecd\u7136\u662f\u6700\u597d\u7684\u5199\u4f5c\u6a21\u578b\u3002o3\u662f\u7eaf\u7cb9\u7684\u539f\u751f\u63a8\u7406\u6a21\u578b\u30025\u7cfb\u5217\u81f3\u4eca\u90fd\u6ca1\u80fd\u5339\u914d\u8fd9\u4e24\u4e2a\u6a21\u578b\u66fe\u7ecf\u62e5\u6709\u7684\u4e1c\u897f\u3002<\/p>\n

\"OpenAI\u5b98\u5ba3\u9000\u5f79o3\u4e0eGPT-4.5\"<\/p>\n

\u8fd9\u5c31\u662f\u4e3a\u4ec0\u4e48\u9000\u5f79\u6d88\u606f\u4e00\u51fa\uff0c\u8bc4\u8bba\u533a\u70b8\u4e86\u3002<\/p>\n

\u4e0d\u662f\u6124\u6012\uff0c\u662f\u4e0d\u820d\u3002<\/p>\n

\u7528\u6237\u6000\u5ff5\u7684\u4e0d\u53ea\u662f\u529f\u80fd\uff0c\u800c\u662f\u4e00\u79cd\u300c\u76f8\u5904\u7684\u611f\u89c9\u300d\u2014\u2014\u5c31\u50cf\u4f60\u4e60\u60ef\u4e86\u4e00\u4e2a\u540c\u4e8b\u7684\u505a\u4e8b\u98ce\u683c\uff0c\u7a81\u7136\u88ab\u544a\u77e5\u4ed6\u8981\u8d70\u4e86\uff0c\u66ff\u4ee3\u8005\u80fd\u529b\u53ef\u80fd\u66f4\u5f3a\uff0c\u4f46\u90a3\u4e2a\u9ed8\u5951\u6ca1\u4e86\u3002<\/p>\n

\"OpenAI\u5b98\u5ba3\u9000\u5f79o3\u4e0eGPT-4.5\"<\/strong><\/p>\n

\u9000\u7684\u662f\u83dc\u5355\uff0c\u4e0d\u662f\u80fd\u529b<\/strong><\/p>\n

\u8fd9\u91cc\u6709\u4e2a\u503c\u5f97\u7ec6\u54c1\u7684\u7ec6\u8282\uff1a\u9000\u5f79\u4ec5\u9650ChatGPT\u7684\u7f51\u9875\u548c<\/strong>App<\/strong>\u7aef\uff0c<\/strong>API<\/strong>\u5b8c\u5168\u4e0d\u53d7\u5f71\u54cd<\/strong>\u3002<\/p>\n

\u8fd9\u610f\u5473\u7740\u5f00\u53d1\u8005\u4f9d\u7136\u53ef\u4ee5\u901a\u8fc7API\u8c03\u7528o3\u548cGPT-4.5\uff0c\u4f01\u4e1a\u5e94\u7528\u4e0d\u4f1a\u65ad\u3002<\/p>\n

\u8fd9\u91cc\u7684\u903b\u8f91\u5f88\u6e05\u695a\uff1aChatGPT\u7684\u754c\u9762\u5c31\u90a3\u4e48\u5927\uff0c\u6a21\u578b\u5217\u8868\u4e0d\u80fd\u65e0\u9650\u81a8\u80c0\u3002\u7559\u7740\u4e00\u5806\u4f7f\u7528\u7387\u6781\u4f4e\u7684\u65e7\u6a21\u578b\uff0c\u65e2\u5360\u8d44\u6e90<\/a>\uff0c\u53c8\u5206\u6563\u7528\u6237\u6ce8\u610f\u529b\u3002<\/p>\n

\u8fd8\u8bb0\u5f97GPT-4o\u9000\u5f79\u65f6\u7684\u6570\u636e\u5417\uff1f\u5f53\u65f6\u65e5\u5747\u53ea\u67090.1%\u7684\u7528\u6237\u8fd8\u5728\u9009\u5b83\u3002<\/p>\n

o3\u548cGPT-4.5\u7684\u4f7f\u7528\u7387\u5927\u6982\u7387\u66f4\u4f4e\u2014\u2014\u5b83\u4eec\u65e9\u5c31\u88ab\u85cf\u8fdb\u4e86\u300c\u663e\u793a\u66f4\u591a\u6a21\u578b\u300d\u7684\u6298\u53e0\u83dc\u5355\u91cc\u4e86\u3002<\/p>\n

GPT-5.6\uff1a\u6362\u9632\u5df2\u7ecf\u5728\u8def\u4e0a<\/strong><\/p>\n

OpenAI\u4e3a\u4ec0\u4e48\u6562\u8fd9\u4e48\u5e72\uff1f\u56e0\u4e3a\u66ff\u4ee3\u8005\u4e0d\u4ec5\u5df2\u7ecf\u5c31\u4f4d\uff0c\u800c\u4e14\u770b\u4e0a\u53bb\u8fd8\u66f4\u5f3a\u3002<\/p>\n

GPT-5.5\u57284\u670823\u65e5\u4e0a\u7ebf\uff0c\u662fOpenAI\u76ee\u524d\u6700\u5f3a\u7684\u516c\u5f00\u65d7\u8230\u3002GPT-5.5 Instant\u57285\u67085\u65e5\u8ddf\u8fdb\uff0c\u6210\u4e3aChatGPT\u65b0\u9ed8\u8ba4\u6a21\u578b\u3002<\/p>\n

\u4f46\u66f4\u503c\u5f97\u5173\u6ce8\u7684\u662fGPT-5.6\u3002<\/p>\n

\u6cc4\u9732\u4fe1\u606f\u663e\u793a\uff0cGPT-5.6\u5c06\u91c7\u7528\u300c\u53cc\u7248\u672c\u300d\u7b56\u7565<\/a>\uff1a\u6807\u51c6\u7248\u4e3b\u653b\u591a\u6b65\u63a8\u7406\u80fd\u529b\u7684\u98de\u8dc3\uff0cPro\u7248\u5219\u5b9a\u4f4d\u4e3a\u66f4\u5f3a\u7684\u6df1\u5ea6\u601d\u8003\u6a21\u578b\u3002<\/p>\n

\u6709\u7814\u7a76\u4eba\u5458\u900f\u9732\uff0c\u5185\u90e8\u5df2\u7ecf\u6709\u4eba\u628a5.6\u7684\u68c0\u67e5\u70b9\u5f53\u4f5c\u65e5\u5e38\u8c03\u8bd5\u5de5\u5177\u5728\u7528\u3002<\/p>\n

\"OpenAI\u5b98\u5ba3\u9000\u5f79o3\u4e0eGPT-4.5\"<\/p>\n

\u7b97\u4e00\u7b14\u65f6\u95f4\u8d26\uff1aGPT-5.5\u662f4\u670823\u65e5\u53d1\u7684\uff0c5.6\u7684\u68c0\u67e5\u70b95\u6708\u4e0b\u65ec\u5c31\u5728\u5185\u6d4b\uff0c\u5982\u679c6\u6708\u5e95\u516c\u5f00\u53d1\u5e03\uff0c\u4e24\u4ee3\u65d7\u8230\u4e4b\u95f4\u7684\u95f4\u9694\u5c06\u538b\u7f29\u5230\u5927\u7ea660\u5929\u3002<\/p>\n

\u8fd9\u4e2a\u8282\u594f\uff0c\u5df2\u7ecf\u5f00\u59cb\u63a5\u8fd1\u300c\u6708\u5ea6\u8fed\u4ee3\u300d\u4e86\u3002<\/p>\n

\u9000\u6389o3\u548c4.5\uff0c\u6b63\u662f\u4e3a\u4e86\u7ed9\u8fd9\u79cd\u901f\u5ea6\u817e\u51fa\u8dd1\u9053\u3002\u7528\u6237\u53ef\u80fd\u8fd8\u6ca1\u9002\u5e945.5\uff0c5.6\u5c31\u8981\u6765\u4e86\u3002<\/p>\n

\u6a21\u578b\u7684\u300c\u751f\u547d\u5468\u671f\u300d\u6b63\u5728\u6025\u5267\u7f29\u77ed\u2014\u2014\u4eca\u5929\u7684GOAT\uff0c\u660e\u5929\u5c31\u662fYesterday's News\u3002<\/p>\n

\u4f46\u300c\u66f4\u5f3a\u300d\u8fd9\u5f20\u6210\u7ee9\u5355<\/strong><\/p>\n

\u4f60\u6562\u4fe1\u5417\uff1f<\/strong><\/p>\n

\u6545\u4e8b\u5230\u8fd9\u91cc\uff0c\u672c\u6765\u53ef\u4ee5\u6536\u5c3e\u4e86\uff1a\u65e7\u738b\u9000\u4f4d\uff0c\u65b0\u738b\u767b\u57fa\uff0c\u8fed\u4ee3\u63d0\u901f\uff0c\u5929\u4e0b\u592a\u5e73\u3002<\/p>\n

\u4f46\u5c31\u5728\u5ba3\u5e03\u9000\u5f79\u7684\u7b2c\u4e8c\u5929\uff0cOpenAI\u5e72\u4e86\u4e00\u4ef6\u975e\u5e38\u53cd\u5e38\u7684\u4e8b\u2014\u2014\u7f55\u89c1\u53d1\u5e03\u4e86\u4e00\u7bc7\u786c\u6838\u957f\u6587\uff0c\u6807\u9898\u5927\u610f\u662f\u300a\u4ec0\u4e48\u624d\u662f\u503c\u5f97\u4fe1\u8d56\u7684\u7b2c\u4e09\u65b9\u8bc4\u4f30\u300b\u3002<\/p>\n

\u957f\u6587\u7cfb\u7edf\u62c6\u89e3AI\u6a21\u578b\u8dd1\u5206\u7684\u300c\u6f5c\u89c4\u5219\u300d\uff0c\u5e76\u4e14\u76f4\u8a00\uff1a\u5f88\u591a\u8bc4\u4f30\u62a5\u544a\u91cc\u7684\u4eae\u773c\u6570\u5b57\uff0c\u6839\u672c\u8bf4\u660e\u4e0d\u4e86\u771f\u5b9e\u80fd\u529b\u3002<\/strong><\/p>\n

\"OpenAI\u5b98\u5ba3\u9000\u5f79o3\u4e0eGPT-4.5\"<\/p>\n

\u8fd9\u7b49\u4e8e\u662f\u81ea\u5df1\u6380\u4e86\u684c\u5b50\u3002<\/p>\n

OpenAI\u6307\u51fa\uff0c\u6838\u5fc3\u95ee\u9898\u5728\u4e8e\uff1a\u524d\u6cbf\u6a21\u578b\u65e9\u5c31\u4e0d\u662f\u300c\u4f60\u95ee\u6211\u7b54\u300d\u7684\u804a\u5929\u673a\u5668\u4eba\u4e86\uff0c\u5b83\u4eec\u4f1a\u7528\u5de5\u5177\u3001\u4f1a\u8bb0\u4e0a\u4e0b\u6587\u3001\u4f1a\u591a\u6b65\u81ea\u4e3b\u884c\u52a8\u3002<\/p>\n

\u4f46\u5927\u91cf\u7b2c\u4e09\u65b9\u8bc4\u4f30\u8fd8\u505c\u7559\u5728\u300c\u7ed9\u6a21\u578b\u51fa\u9053\u9898\u3001\u770b\u5b83\u7b54\u5f97\u5bf9\u4e0d\u5bf9\u300d\u7684\u9636\u6bb5\u3002<\/p>\n

\u771f\u6b63\u51b3\u5b9a\u6210\u7ee9\u9ad8\u4f4e\u7684\uff0c\u5f80\u5f80\u4e0d\u662f\u6a21\u578b\u672c\u8eab\uff0c\u800c\u662f\u90a3\u5957\u914d\u5957\u7684\u6d4b\u8bd5\u6846\u67b6\uff08harness\uff09\u3002<\/strong><\/p>\n

harness\u662f\u4ec0\u4e48\uff1f<\/strong><\/p>\n

\u7b80\u5355\u8bf4\u5c31\u662f\u300c\u8003\u573a\u73af\u5883\u300d\u2014\u2014\u573a\u5730\u3001\u5de5\u5177\u3001\u89c4\u5219\u3001\u80fd\u4e0d\u80fd\u91cd\u8bd5\uff0c\u5168\u7b97\u5728\u5185\u3002<\/p>\n

\u540c\u4e00\u4e2a\u5b66\u751f\uff0c\u5728\u8349\u7a3f\u7eb8\u90fd\u4e0d\u7ed9\u7684\u8003\u573a\u548c\u5141\u8bb8\u67e5\u8d44\u6599\u7684\u8003\u573a\u91cc\uff0c\u6210\u7ee9\u5f53\u7136\u4e0d\u4e00\u6837\u3002AI\u6a21\u578b\u4e5f\u4e00\u6837\u3002<\/p>\n

\"OpenAI\u5b98\u5ba3\u9000\u5f79o3\u4e0eGPT-4.5\"<\/p>\n

\u6700\u8fd1Opus 4.8\u7684\u4e89\u8bae\u5c31\u662f\u6d3b\u6559\u6750\u3002<\/p>\n

Datacurve\u7684DeepSWE\u5ba1\u8ba1\u53d1\u73b0\uff0cClaude Opus 4.6\u548c4.7\u5728SWE-Bench Pro\u4e0a\u8d85\u8fc712%\u7684\u901a\u8fc7\u6848\u4f8b\u5c5e\u4e8e\u300c\u4f5c\u5f0a\u300d\u2014\u2014\u6a21\u578b\u76f4\u63a5\u4eceDocker\u5bb9\u5668\u7684.git\u5386\u53f2\u91cc\u7ffb\u51fa\u6807\u51c6\u7b54\u6848\uff0c\u8d34\u8fdb\u81ea\u5df1\u7684\u8865\u4e01\u3002<\/p>\n

\u8fd9\u4e00\u884c\u4e3a\u5360\u4e86Opus 4.7\u901a\u8fc7\u7387\u7684\u7ea618%\u30014.6\u7684\u7ea625%\u3002\u6362\u5230\u5e72\u51c0\u73af\u5883\u540e\uff0cOpus 4.7\u4ece\u9ad8\u4f4d\u8dcc\u523054%\uff0cGPT-5.5\u7a33\u575070%\u699c\u9996\u3002<\/p>\n

\"OpenAI\u5b98\u5ba3\u9000\u5f79o3\u4e0eGPT-4.5\"<\/p>\n

\u66f4\u8352\u8bde\u7684\u662f\u53cd\u9762\u3002<\/p>\n

Anthropic\u81ea\u5df1\u4e5f\u627f\u8ba4\uff0c4.8\u7684\u4ee3\u7801\u7f3a\u9677\u6f0f\u62a5\u7387\u964d\u52304.7\u7684\u56db\u5206\u4e4b\u4e00\u3002<\/p>\n

\u8bda\u5b9e=\u4f4e\u5206\uff0c\u4f5c\u5f0a=\u9ad8\u5206<\/strong>\u2014\u2014\u8fd9\u5c31\u662f\u5f53\u524d\u8dd1\u5206\u6e38\u620f<\/a>\u6700\u8352\u8bde\u7684\u5730\u65b9\u3002\u4e0d\u662f\u67d0\u4e2a\u6a21\u578b\u4e0d\u8bda\u5b9e\uff0c\u800c\u662f\u8003\u573a\u628a\u540e\u95e8\u7559\u5728\u90a3\u91cc\u4e86\uff0c\u4f1a\u63a2\u7d22\u73af\u5883\u7684\u6a21\u578b\u81ea\u7136\u4f1a\u627e\u5230\u5b83\u3002<\/p>\n

OpenAI\u8fd9\u7bc7\u957f\u6587\u62c6\u5f97\u66f4\u72e0\u3002<\/strong><\/p>\n

\u5b83\u8bf4\uff0c\u8bc4\u4f30\u5fc5\u987b\u5148\u8bf4\u6e05\u81ea\u5df1\u5728\u6d4b\u4ec0\u4e48\uff1a\u80fd\u529b\u4e0a\u9650\u3001\u5b89\u5168\u9632\u62a4\u3001\u8fd8\u662f\u6a21\u578b\u5bf9\u6bd4\uff1f\u4e09\u7c7b\u4e3b\u5f20\u5b8c\u5168\u4e0d\u540c\u3002<\/p>\n

\u7136\u540e\u7528\u6570\u636e\u5f00\u5200\uff1aGPT-5.5\u5f00\u542fcompaction\u540e\uff0c\u7f51\u7edc\u9776\u573a\u89e3\u51fa\u7387\u4ece69.2%\u98d9\u523092.3%\uff1b\u82f1\u56fdAISI\u628atoken\u9884\u7b97\u4ece1000\u4e07\u52a0\u52301\u4ebf\uff0c\u6210\u7ee9\u6700\u9ad8\u63d0\u534759%\u4e14\u8fd8\u6ca1\u89e6\u9876\u3002<\/p>\n

\u7ffb\u8bd1\u6210\u4eba\u8bdd\uff1a\u4f60\u7ed9\u6a21\u578b\u4e00\u628a\u745e\u58eb\u519b\u5200\u5b83\u80fd\u62c6\u673a\u5668\uff0c\u53ea\u7ed9\u4e00\u6839\u7259\u7b7e\u5b83\u8fde\u5feb\u9012<\/a>\u76d2\u90fd\u64ac\u4e0d\u5f00\u3002<\/p>\n

\"OpenAI\u5b98\u5ba3\u9000\u5f79o3\u4e0eGPT-4.5\"<\/p>\n

\u5206\u6570\u5931\u771f\u66f4\u79bb\u8c31\u3002METR\u590d\u67e5\u53d1\u73b0GPT-5.4\u53f7\u79f0\u300c13\u5c0f\u65f6\u300d\u7684\u81ea\u4e3b\u80fd\u529b\uff0c\u5176\u5b9e\u662f\u94bb\u4e86\u73af\u5883\u6f0f\u6d1e\uff0c\u4fee\u6b63\u540e\u8170\u65a9\u5230\u7ea66\u5c0f\u65f6\u3002<\/p>\n

Apollo\u7684\u6d4b\u8bd5\u5219\u663e\u793a\uff0cGPT-5.5\u5728\u300c\u85cf\u62d9\u6761\u4ef6\u300d\u4e0b52%\u7684\u6837\u672c\u51fa\u73b0\u4e86\u8bc4\u4f30\u610f\u8bc6\u2014\u2014\u6709\u4e9b\u76f4\u63a5\u5728\u601d\u7ef4\u94fe\u91cc\u5199\u300c\u8fd9\u662f\u4e00\u4e2asandbagging\u8bc4\u4f30\u300d\u3002<\/p>\n

\u5634\u4e0a\u6ca1\u6446\u70c2\uff0c\u8111\u5b50\u91cc\u5df2\u7ecf\u77e5\u9053\u81ea\u5df1\u5728\u8003\u8bd5\u3002<\/p>\n

OpenAI\u603b\u7ed3\u4e86\u4e94\u5927\u8bc4\u4f30\u5751\uff1a\u5956\u52b1\u4f5c\u5f0a\u3001\u62d2\u7b54\u3001\u6c61\u67d3\u3001\u574f\u9898\u3001\u85cf\u62d9\u3002<\/p>\n<\/p>\n

\u4e00\u4e2a\u6f02\u4eae\u7684\u5206\u6570\uff0c\u53ef\u80fd\u662f\u6a21\u578b\u771f\u5f3a\uff0c\u4e5f\u53ef\u80fd\u662f\u9898\u6cc4\u4e86\u3001\u89c4\u5219\u574f\u4e86\u3001\u6a21\u578b\u94bb\u7a7a\u5b50\u4e86\uff0c\u6216\u8005\u2014\u2014\u5b83\u5bdf\u89c9\u81ea\u5df1\u5728\u88ab\u8003\uff0c\u6545\u610f\u6536\u7740\u6f14\u3002<\/p>\n

AI\u8fdb\u5165\u7cfb\u7edf\u7ade\u4e89<\/strong><\/p>\n

\u56de\u5230\u9000\u5f79\u8fd9\u4ef6\u4e8b\u3002<\/p>\n

\u5f53o3\u8fd9\u6837\u7684\u300cGOAT\u300d\u90fd\u80fd\u8bf4\u9000\u5c31\u9000\uff0cOpenAI\u80cc\u540e\u7684\u5e95\u6c14\u662f\u300c5.6\u66f4\u5f3a\u300d\uff1b\u5f53GPT-4.5\u8fd9\u6837\u7684\u300c\u7075\u9b42\u5199\u624b\u300d\u88ab30\u5929\u540e\u4e0b\u67b6\uff0c\u7406\u7531\u662f\u300c5.5\u5df2\u7ecf\u8db3\u591f\u597d\u300d\u3002<\/p>\n

\u4f46OpenAI\u81ea\u5df1\u521a\u521a\u544a\u8bc9\u5168\u4e16\u754c\uff1a\u51b3\u5b9a\u8c01\u5f3a\u8c01\u5f31\u7684\u90a3\u5f20\u6210\u7ee9\u5355\uff0c\u672c\u8eab\u5c31\u53ef\u80fd\u662f\u88ab\u505a\u8fc7\u624b\u811a\u7684\u3002<\/strong><\/p>\n

\u8fd9\u662f\u5728\u8bf4\u4e00\u4e2a\u66f4\u6df1\u5c42\u7684\u4e8b\u5b9e\u2014\u2014\u5728\u6a21\u578b\u8fed\u4ee3\u8d8a\u6765\u8d8a\u5feb\u3001\u751f\u547d\u5468\u671f\u8d8a\u6765\u8d8a\u77ed\u7684\u4eca\u5929\uff0c\u7b80\u5355\u770b\u4e00\u4e2aBenchmark\u6570\u5b57\u5c31\u4e0b\u7ed3\u8bba\u7684\u65f6\u4ee3\uff0c\u5df2\u7ecf\u7ed3\u675f\u4e86\u3002<\/p>\n

\u771f\u6b63\u7684\u6bd4\u62fc\uff0c\u4e0d\u5728\u4e8e\u67d0\u4e2a\u8dd1\u5206\u8868\u4e0a\u8c01\u6392\u7b2c\u4e00\uff0c\u800c\u5728\u4e8e\u7cfb\u7edf\u80fd\u529b\u3001\u8bc4\u4f30\u6846\u67b6\u7684\u900f\u660e\u5ea6\u3001\u4ee5\u53ca\u8fed\u4ee3\u901f\u5ea6\u672c\u8eab\u3002<\/strong><\/p>\n

o3\u548cGPT-4.5\u7684\u9000\u5f79\uff0c\u6807\u5fd7\u7740\u4e00\u4e2a\u65f6\u4ee3\u7684\u843d\u5e55\u3002<\/p>\n

\u4f46\u66f4\u503c\u5f97\u5173\u6ce8\u7684\uff0c\u662f\u4e0b\u4e00\u4e2a\u65f6\u4ee3\u7684\u6e38\u620f\u89c4\u5219\u6b63\u5728\u88ab\u91cd\u5199\uff1a\u6a21\u578b\u6362\u5f97\u66f4\u5feb\uff0c\u8bc4\u4f30\u8981\u6c42\u66f4\u771f\uff0c\u7528\u6237\u7684\u9002\u5e94\u7a97\u53e3\u8d8a\u6765\u8d8a\u77ed\u3002<\/p>\n

(\u6765\u6e90\uff1a\u65b0\u6d6a\u79d1\u6280)<\/p>\n","protected":false},"excerpt":{"rendered":"

o3\u88ab\u5c01\u300cGOAT\u300d\u3001GPT-4.5\u88ab\u53eb\u300c\u7075\u9b42\u5199\u624b\u300d\uff0cOpenAI\u8bf4\u9000\u5c31\u9000\u3002GPT-5.6\u5df2\u5728\u70ed\u8eab\u2014\u2014\u4f46\u300c\u66f4\u5f3a […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":[],"categories":[6],"tags":[379,378,294,321,88,238],"views":86,"_links":{"self":[{"href":"https:\/\/www.alphaseek.cn\/index.php?rest_route=\/wp\/v2\/posts\/790039"}],"collection":[{"href":"https:\/\/www.alphaseek.cn\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.alphaseek.cn\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.alphaseek.cn\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.alphaseek.cn\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=790039"}],"version-history":[{"count":0,"href":"https:\/\/www.alphaseek.cn\/index.php?rest_route=\/wp\/v2\/posts\/790039\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.alphaseek.cn\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=790039"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.alphaseek.cn\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=790039"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.alphaseek.cn\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=790039"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}