{"id":28158,"date":"2025-02-06T11:27:38","date_gmt":"2025-02-06T03:27:38","guid":{"rendered":"https:\/\/www.1ai.net\/?p=28158"},"modified":"2025-02-06T11:31:08","modified_gmt":"2025-02-06T03:31:08","slug":"%e8%ae%ad%e7%bb%83%e6%88%90%e6%9c%ac%e4%b8%8d%e5%88%b0-50-%e7%be%8e%e5%85%83%ef%bc%8c%e7%a0%94%e7%a9%b6%e4%ba%ba%e5%91%98%e6%89%93%e9%80%a0%e5%87%ba%e5%aa%b2%e7%be%8e-openai-o1-%e7%9a%84%e6%8e%a8","status":"publish","type":"post","link":"https:\/\/www.1ai.net\/en\/28158.html","title":{"rendered":"For Less Than $50 to Train, Researchers Build an Inference Model That Rivals OpenAI o1"},"content":{"rendered":"

2 \u6708 6 \u65e5\u6d88\u606f\uff0c\u4e00\u4efd\u4e0a\u5468\u4e94\u53d1\u5e03\u7684\u7814\u7a76\u62a5\u544a\u663e\u793a\uff0c\u65af\u5766\u798f\u5927\u5b66\u548c\u534e\u76db\u987f\u5927\u5b66\u7684\u4eba\u5de5\u667a\u80fd\u7814\u7a76\u4eba\u5458\u4ec5\u82b1\u8d39\u4e0d\u5230 50 \u7f8e\u5143\uff08\u6ce8\uff1a\u5f53\u524d\u7ea6 364 \u5143\u4eba\u6c11\u5e01\uff09\u7684\u4e91\u8ba1\u7b97\u8d39\u7528\uff0c\u5c31\u6210\u529f\u8bad\u7ec3\u51fa\u4e00\u4e2a\u5177\u5907\u201c\u63a8\u7406\u201d\u80fd\u529b\u7684\u4eba\u5de5\u667a\u80fd\u6a21\u578b<\/a>\u3002<\/p>\n

\"b44fec88j00sr8s8u0030d000v900b8p\"<\/p>\n

\u8be5\u6a21\u578b\u540d\u4e3a s1\uff0c\u5728\u6570\u5b66\u548c\u7f16\u7a0b\u80fd\u529b\u6d4b\u8bd5\u4e2d\u8868\u73b0\u51fa\u4e0e OpenAI \u7684 o1 \u548c DeepSeek \u7684 r1 \u7b49\u9876\u5c16\u63a8\u7406\u6a21\u578b<\/a>\u76f8\u4f3c\u7684\u6c34\u5e73<\/strong>\u3002\u76ee\u524d\uff0cs1 \u6a21\u578b\u53ca\u5176\u8bad\u7ec3\u6240\u7528\u7684\u6570\u636e\u548c\u4ee3\u7801\u5df2\u5728 GitHub \u4e0a\u5f00\u6e90<\/a>\u3002<\/p>\n

s1 \u56e2\u961f\u8868\u793a\uff0c\u4ed6\u4eec\u901a\u8fc7\u201c\u84b8\u998f\u201d\u6280\u672f\u521b\u5efa\u4e86\u8be5\u4eba\u5de5\u667a\u80fd\u6a21\u578b<\/strong>\uff0c\u8be5\u6280\u672f\u65e8\u5728\u901a\u8fc7\u8bad\u7ec3\u6a21\u578b\u6765\u5b66\u4e60\u53e6\u4e00\u4e2a\u4eba\u5de5\u667a\u80fd\u6a21\u578b\u7684\u7b54\u6848\uff0c\u4ece\u800c\u63d0\u53d6\u5176\u201c\u63a8\u7406\u201d\u80fd\u529b\u3002\u7814\u7a76\u4eba\u5458\u900f\u9732\uff0cs1 \u662f\u4ece\u8c37\u6b4c\u7684\u63a8\u7406\u6a21\u578b Gemini 2.0 Flash Thinking Experimental \u4e2d\u84b8\u998f\u51fa\u6765\u7684\u3002\u4e0a\u4e2a\u6708\uff0c\u52a0\u5dde\u5927\u5b66\u4f2f\u514b\u5229\u5206\u6821\u7684\u7814\u7a76\u4eba\u5458\u4e5f\u4f7f\u7528\u4e86\u76f8\u540c\u7684\u84b8\u998f\u65b9\u6cd5\uff0c\u4ee5\u7ea6 450 \u7f8e\u5143\u7684\u6210\u672c\u521b\u5efa\u4e86\u4e00\u4e2a\u4eba\u5de5\u667a\u80fd\u63a8\u7406\u6a21\u578b\u3002<\/p>\n

\u7c7b\u4f3c s1 \u6a21\u578b\u7684\u51fa\u73b0\u4e5f\u5f15\u53d1\u4e86\u5173\u4e8e\u4eba\u5de5\u667a\u80fd\u6a21\u578b\u5546\u54c1\u5316\u7684\u95ee\u9898 \u2014\u2014 \u5982\u679c\u6709\u4eba\u53ef\u4ee5\u7528\u76f8\u5bf9\u8f83\u4f4e\u7684\u6210\u672c\u5c31\u80fd\u590d\u5236\u4e00\u4e2a\u4ef7\u503c\u6570\u767e\u4e07\u7f8e\u5143\u7684\u6a21\u578b\uff0c\u90a3\u4e48\u5927\u578b\u79d1\u6280\u516c\u53f8\u7684\u201c\u62a4\u57ce\u6cb3\u201d\u5728\u54ea\u91cc\u5462\uff1f<\/p>\n

\u4e0d\u51fa\u6240\u6599\uff0c\u5927\u578b\u4eba\u5de5\u667a\u80fd\u5b9e\u9a8c\u5ba4\u5bf9\u6b64\u5e76\u4e0d\u6ee1\u610f\uff0c\u4f8b\u5982 OpenAI \u6b64\u524d\u5c31\u6307\u8d23 DeepSeek \u4e0d\u5f53\u83b7\u53d6\u5176 API \u6570\u636e\u7528\u4e8e\u6a21\u578b\u84b8\u998f\u3002<\/p>\n

s1 \u7684\u7814\u7a76\u4eba\u5458\u5e0c\u671b\u627e\u5230\u5b9e\u73b0\u5f3a\u5927\u63a8\u7406\u6027\u80fd\u548c\u201c\u6d4b\u8bd5\u65f6\u6269\u5c55\u201d\uff08\u5373\u5141\u8bb8\u4eba\u5de5\u667a\u80fd\u6a21\u578b\u5728\u56de\u7b54\u95ee\u9898\u4e4b\u524d\u8fdb\u884c\u66f4\u591a\u601d\u8003\uff09\u7684\u6700\u7b80\u5355\u65b9\u6cd5\uff0c\u8fd9\u4e9b\u662f OpenAI \u7684 o1 \u4e2d\u7684\u4e00\u4e9b\u7a81\u7834\u3002<\/p>\n

s1 \u7684\u8bba\u6587\u8868\u660e\uff0c\u53ef\u4ee5\u4f7f\u7528\u4e00\u79cd\u79f0\u4e3a\u76d1\u7763\u5fae\u8c03\uff08SFT\uff09\u7684\u65b9\u6cd5\uff0c\u53ef\u4ee5\u4f7f\u7528\u76f8\u5bf9\u8f83\u5c0f\u7684\u6570\u636e\u96c6\u6765\u84b8\u998f\u63a8\u7406\u6a21\u578b<\/strong>\u3002\u5728 SFT \u4e2d\uff0c\u4eba\u5de5\u667a\u80fd\u6a21\u578b\u4f1a\u88ab\u660e\u786e\u6307\u793a\u5728\u6570\u636e\u96c6\u4e2d\u6a21\u4eff\u67d0\u4e9b\u884c\u4e3a\u3002SFT \u6bd4 DeepSeek \u7528\u4e8e\u8bad\u7ec3\u5176 R1 \u6a21\u578b\u7684\u5927\u89c4\u6a21\u5f3a\u5316\u5b66\u4e60\u65b9\u6cd5\u66f4\u5177\u6210\u672c\u6548\u76ca\u3002<\/p>\n

\u8c37\u6b4c\u901a\u8fc7\u5176 Google AI Studio \u5e73\u53f0\u514d\u8d39\u63d0\u4f9b Gemini 2.0 Flash Thinking Experimental \u6a21\u578b\u7684\u8bbf\u95ee\u6743\u9650\uff0c\u4f46\u6bcf\u5929\u6709\u4f7f\u7528\u9650\u5236\u3002\u7136\u800c\uff0c\u5176\u6761\u6b3e\u7981\u6b62\u5bf9\u6a21\u578b\u8fdb\u884c\u9006\u5411\u5de5\u7a0b\uff0c\u4ee5\u5f00\u53d1\u4e0e\u8c37\u6b4c\u81ea\u8eab\u4eba\u5de5\u667a\u80fd\u4ea7\u54c1\u7ade\u4e89\u7684\u670d\u52a1\u3002<\/p>\n

S1 \u57fa\u4e8e\u963f\u91cc\u5df4\u5df4\u65d7\u4e0b\u4e2d\u56fd\u4eba\u5de5\u667a\u80fd\u5b9e\u9a8c\u5ba4 Qwen \u63d0\u4f9b\u7684\u4e00\u6b3e\u5c0f\u578b\u3001\u73b0\u6210\u7684\u514d\u8d39\u4eba\u5de5\u667a\u80fd\u6a21\u578b\u3002\u4e3a\u4e86\u8bad\u7ec3 s1\uff0c\u7814\u7a76\u4eba\u5458\u521b\u5efa\u4e86\u4e00\u4e2a\u4ec5\u5305\u542b 1000 \u4e2a\u7cbe\u5fc3\u7b56\u5212\u7684\u95ee\u9898\u7684\u6570\u636e\u96c6\uff0c\u4ee5\u53ca\u8fd9\u4e9b\u95ee\u9898\u7684\u7b54\u6848\uff0c\u4ee5\u53ca\u8c37\u6b4c Gemini 2.0 Flash Thinking Experimental \u7ed9\u51fa\u7684\u6bcf\u4e2a\u7b54\u6848\u80cc\u540e\u7684\u201c\u601d\u8003\u201d\u8fc7\u7a0b\u3002<\/strong><\/p>\n

\u7814\u7a76\u4eba\u5458\u8868\u793a\uff0c\u5728\u8bad\u7ec3 s1 \u540e\uff08\u4f7f\u7528 16 \u4e2a Nvidia H100 GPU\uff0c\u8017\u65f6\u4e0d\u5230 30 \u5206\u949f\uff09<\/strong>\uff0cs1 \u5728\u67d0\u4e9b\u4eba\u5de5\u667a\u80fd\u57fa\u51c6\u6d4b\u8bd5\u4e2d\u53d6\u5f97\u4e86\u826f\u597d\u7684\u8868\u73b0\u3002\u53c2\u4e0e\u8be5\u9879\u76ee\u7684\u65af\u5766\u798f\u5927\u5b66\u7814\u7a76\u5458 Niklas Muennighoff \u544a\u8bc9 TechCrunch\uff0c\u76ee\u524d\u79df\u7528\u8fd9\u4e9b\u8ba1\u7b97\u8d44\u6e90\u7684\u6210\u672c\u7ea6\u4e3a 20 \u7f8e\u5143\u3002<\/p>\n

\u7814\u7a76\u4eba\u5458\u4f7f\u7528\u4e86\u4e00\u4e2a\u5de7\u5999\u7684\u6280\u5de7\u6765\u8ba9 s1 \u68c0\u67e5\u5176\u5de5\u4f5c\u5e76\u5ef6\u957f\u5176\u201c\u601d\u8003\u201d\u65f6\u95f4\uff1a\u4ed6\u4eec\u8ba9\u5b83\u201c\u7b49\u5f85\u201d\u3002\u8bba\u6587\u663e\u793a\uff0c\u5728 s1 \u7684\u63a8\u7406\u8fc7\u7a0b\u4e2d\u6dfb\u52a0\u201c\u7b49\u5f85\u201d\u4e00\u8bcd\uff0c\u6709\u52a9\u4e8e\u6a21\u578b\u83b7\u5f97\u7a0d\u5fae\u66f4\u51c6\u786e\u7684\u7b54\u6848\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"

Feb. 6 (Bloomberg) -- Artificial intelligence researchers at Stanford University and the University of Washington have successfully trained an AI model with \"reasoning\" capabilities for less than $50 (note: currently about 364 yuan) in cloud computing, according to a study released on Friday. The model, called s1, performed similarly to top reasoning models such as OpenAI's o1 and DeepSeek's r1 in tests of mathematical and programming ability. The s1 model and the data and code used for its training are currently open source on GitHub. The s1 team said they created the AI model through a \"distillation\" technique, which is designed to train a model to learn the answers of another AI model.<\/p>","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[146],"tags":[599,5023],"collection":[],"class_list":["post-28158","post","type-post","status-publish","format-standard","hentry","category-news","tag-599","tag-5023"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/28158","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/comments?post=28158"}],"version-history":[{"count":0,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/28158\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/media?parent=28158"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/categories?post=28158"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/tags?post=28158"},{"taxonomy":"collection","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/collection?post=28158"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}