{"id":36,"date":"2025-01-27T21:46:30","date_gmt":"2025-01-27T13:46:30","guid":{"rendered":"http:\/\/jp5.ssqwert.eu.org\/?p=36"},"modified":"2025-01-27T21:46:30","modified_gmt":"2025-01-27T13:46:30","slug":"%e6%9c%ac%e5%9c%b0%e9%83%a8%e7%bd%b2-deepseek-r1-%e5%a4%a7%e6%a8%a1%e5%9e%8b%ef%bc%81%e5%85%8d%e8%b4%b9%e5%bc%80%e6%ba%90%ef%bc%8c%e5%aa%b2%e7%be%8eopenai-o1%e8%83%bd%e5%8a%9b","status":"publish","type":"post","link":"https:\/\/jp5.ssqwert.eu.org\/index.php\/2025\/01\/27\/%e6%9c%ac%e5%9c%b0%e9%83%a8%e7%bd%b2-deepseek-r1-%e5%a4%a7%e6%a8%a1%e5%9e%8b%ef%bc%81%e5%85%8d%e8%b4%b9%e5%bc%80%e6%ba%90%ef%bc%8c%e5%aa%b2%e7%be%8eopenai-o1%e8%83%bd%e5%8a%9b\/","title":{"rendered":"\u672c\u5730\u90e8\u7f72 DeepSeek-R1 \u5927\u6a21\u578b\uff01\u514d\u8d39\u5f00\u6e90\uff0c\u5ab2\u7f8eOpenAI-o1\u80fd\u529b"},"content":{"rendered":"\n

\u6700\u8fd1\uff0c\u4e00\u5bb6\u540d\u53ebDeepSeek\u7684\u521d\u521b\u516c\u53f8\u7ecf\u8fc7\u6280\u672f\u8fed\u4ee3\u4e0e\u5347\u7ea7\uff0c\u53d1\u5e03\u4e86\u5168\u65b0\u4e00\u4ee3\u5927\u6a21\u578b\uff0c\u201cDeepSeek-V3\u201d\u3002\u7531\u4e8e\u8fd9\u6b3e\u5927\u6a21\u578b\u592a\u8fc7\u597d\u7528\uff0cDeepSeek R1 \u66f4\u662f\u76f4\u63a5\u514d\u8d39\u5f00\u6e90\uff0c\u5728AI\u53d1\u70e7\u53cb\u5708\u5b50\u4f20\u64ad\u540e\uff0c\u4f20\u5230\u4e86\u6d77\u5916\u793e\u4ea4\u5e73\u53f0\u3001\u6280\u672f\u8bba\u575b\uff0c\u5f15\u53d1\u4e86\u6d77\u5916\u7f51\u53cb\u7684\u8fde\u8fde\u79f0\u8d5e\u3002<\/p>\n\n\n\n

\"0ca1c0838c20250125140830\"\/<\/figure>\n\n\n\n

\u5404\u9879\u6027\u80fd\u6307\u6807\u66f4\u662f\u548cOpenAI-o1 \u6a21\u578b\u4e0d\u76f8\u4e0a\u4e0b\uff0c\u751a\u81f3\u505a\u5230\u4e86\u5c0f\u90e8\u5206\u7684\u8d85\u8d8a\uff0c\u5173\u952e\u662f\u5f00\u6e90\u7684\uff0c\u6211\u4eec\u53ef\u4ee5\u672c\u5730\u90e8\u7f72\u4f7f\u7528<\/p>\n\n\n\n

\"d41da917e620250125141204\"\/<\/figure>\n\n\n\n

1\u3001\u672c\u5730\u90e8\u7f72\uff0c<\/strong>\u6211\u4eec\u53ef\u4ee5\u901a\u8fc7Ollama\u6765\u8fdb\u884c\u5b89\u88c5<\/strong><\/h2>\n\n\n\n

Ollama \u5b98\u65b9\u7248\uff1a<\/strong>\u3010\u70b9\u51fb\u524d\u5f80<\/a>\u3011<\/h2>\n\n\n\n

Web UI \u63a7\u5236\u7aef\u3010\u70b9\u51fb\u5b89\u88c5<\/a>\u3011<\/strong><\/p>\n\n\n\n

\u5b89\u88c5\u547d\u4ee4<\/p>\n\n\n\n

1.5B Qwen DeepSeek R1<\/strong><\/p>\n\n\n\n

ollama run deepseek-r1:1.5b<\/pre>\n\n\n\n
7B Qwen DeepSeek R1<\/strong><\/p>\n\n\n\n
ollama run deepseek-r1:7b<\/pre>\n\n\n\n
8B Llama DeepSeek R1<\/strong><\/p>\n\n\n\n
ollama run deepseek-r1:8b<\/pre>\n\n\n\n
14B Qwen DeepSeek R1<\/strong><\/p>\n\n\n\n
ollama run deepseek-r1:14b<\/pre>\n\n\n\n
32B Qwen DeepSeek R1<\/strong><\/p>\n\n\n\n
ollama run deepseek-r1:32b<\/pre>\n\n\n\n
70B Llama DeepSeek R1<\/strong><\/p>\n\n\n\n
ollama run deepseek-r1:70b<\/pre>\n\n\n\n
2. \u66f4\u591a\u6a21\u578b\u4e0b\u8f7d<\/h2>\n\n\n\n
DeepSeek-R1<\/h3>\n\n\n\n
\u6a21\u578b<\/strong><\/th>#\u603b\u53c2\u6570<\/strong><\/th>#\u5df2\u6fc0\u6d3b\u53c2\u6570<\/strong><\/th>\u4e0a\u4e0b\u6587\u957f\u5ea6<\/strong><\/th>\u4e0b\u8f7d<\/strong><\/th><\/tr><\/thead>
DeepSeek-R1-Zero<\/td>671B<\/td>37B<\/td>128\u5343<\/td>\ud83e\udd17 HuggingFace<\/a><\/td><\/tr>
DeepSeek-R1<\/td>671B<\/td>37B<\/td>128\u5343<\/td>\ud83e\udd17 HuggingFace<\/a><\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n
DeepSeek-R1-Zero \u548c DeepSeek-R1 \u57fa\u4e8e DeepSeek-V3-Base \u8fdb\u884c\u8bad\u7ec3\u3002\u6709\u5173\u6a21\u578b\u67b6\u6784\u7684\u66f4\u591a\u8be6\u7ec6\u4fe1\u606f\uff0c\u8bf7\u53c2\u9605DeepSeek-V3<\/a>\u5b58\u50a8\u5e93\u3002<\/p>\n\n\n\n
DeepSeek-R1-Distill \u6a21\u578b<\/h3>\n\n\n\n
\u6a21\u578b<\/strong><\/th>\u57fa\u7840\u6a21\u578b<\/strong><\/th>\u4e0b\u8f7d<\/strong><\/th><\/tr><\/thead>
DeepSeek-R1-Distill-Qwen-1.5B<\/td>Qwen2.5-Math-1.5B<\/a><\/td>\ud83e\udd17 HuggingFace<\/a><\/td><\/tr>
DeepSeek-R1-Distill-Qwen-7B<\/td>Qwen2.5-Math-7B<\/a><\/td>\ud83e\udd17 HuggingFace<\/a><\/td><\/tr>
DeepSeek-R1-Distill-Llama-8B<\/td>Llama-3.1-8B<\/a><\/td>\ud83e\udd17 HuggingFace<\/a><\/td><\/tr>
DeepSeek-R1-Distill-Qwen-14B<\/td>Qwen2.5-14B<\/a><\/td>\ud83e\udd17 HuggingFace<\/a><\/td><\/tr>
DeepSeek-R1-Distill-Qwen-32B<\/td>Qwen2.5-32B<\/a><\/td>\ud83e\udd17 HuggingFace<\/a><\/td><\/tr>
DeepSeek-R1-Distill-Llama-70B<\/td>Llama-3.3-70B-Instruct<\/a><\/td>\ud83e\udd17 HuggingFace<\/a><\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n
DeepSeek-R1-Distill \u6a21\u578b\u57fa\u4e8e\u5f00\u6e90\u6a21\u578b\u8fdb\u884c\u4e86\u5fae\u8c03\uff0c\u4f7f\u7528\u4e86 DeepSeek-R1 \u751f\u6210\u7684\u6837\u672c\u3002\u6211\u4eec\u5bf9\u5176\u914d\u7f6e\u548c\u5206\u8bcd\u5668\u8fdb\u884c\u4e86\u8f7b\u5fae\u66f4\u6539\u3002\u8bf7\u4f7f\u7528\u6211\u4eec\u7684\u8bbe\u7f6e\u6765\u8fd0\u884c\u8fd9\u4e9b\u6a21\u578b\u3002<\/p>\n\n\n\n
 <\/h2>\n\n\n\n
4.\u8bc4\u4f30\u7ed3\u679c<\/h2>\n\n\n\n
DeepSeek-R1-\u8bc4\u4f30<\/h3>\n\n\n\n
\u5bf9\u4e8e\u6211\u4eec\u6240\u6709\u7684\u6a21\u578b\uff0c\u6700\u5927\u751f\u6210\u957f\u5ea6\u8bbe\u7f6e\u4e3a 32,768 \u4e2a token\u3002\u5bf9\u4e8e\u9700\u8981\u91c7\u6837\u7684\u57fa\u51c6\uff0c\u6211\u4eec\u4f7f\u7528\u7684\u6e29\u5ea6\u4e3a0.6\uff0ctop-p \u503c\u4e3a0.95\uff0c\u5e76\u4e3a\u6bcf\u4e2a\u67e5\u8be2\u751f\u6210 64 \u4e2a\u54cd\u5e94\u6765\u4f30\u8ba1 pass@1\u3002<\/p>\n\n\n\n
\u7c7b\u522b<\/th>\u57fa\u51c6\uff08\u516c\u5236\uff09<\/th>\u514b\u52b3\u5fb7-3.5-\u5341\u56db\u884c\u8bd7-1022<\/th>GPT-4o 0513<\/th>DeepSeek V3<\/th>OpenAI o1-mini<\/th>OpenAI o1-1217<\/th>DeepSeek R1<\/th><\/tr><\/thead>
 <\/td>\u5efa\u7b51\u5b66<\/td>\u2013<\/td>\u2013<\/td>\u6559\u80b2\u90e8<\/td>\u2013<\/td>\u2013<\/td>\u6559\u80b2\u90e8<\/td><\/tr>
 <\/td># \u6fc0\u6d3b\u53c2\u6570<\/td>\u2013<\/td>\u2013<\/td>37B<\/td>\u2013<\/td>\u2013<\/td>37B<\/td><\/tr>
 <\/td># \u603b\u53c2\u6570<\/td>\u2013<\/td>\u2013<\/td>671B<\/td>\u2013<\/td>\u2013<\/td>671B<\/td><\/tr>
\u82f1\u8bed<\/td>MMLU\uff08\u901a\u8fc7@1\uff09<\/td>88.3<\/td>87.2<\/td>88.5<\/td>85.2<\/td>91.8<\/strong><\/td>90.8<\/td><\/tr>
 <\/td>MMLU-Redux\uff08EM\uff09<\/td>88.9<\/td>88.0<\/td>89.1<\/td>86.7<\/td>\u2013<\/td>92.9<\/strong><\/td><\/tr>
 <\/td>MMLU-Pro\uff08EM\uff09<\/td>78.0<\/td>72.6<\/td>75.9<\/td>80.3<\/td>\u2013<\/td>84.0<\/strong><\/td><\/tr>
 <\/td>\u6389\u843d (3 \u53d1 F1)<\/td>88.3<\/td>83.7<\/td>91.6<\/td>83.9<\/td>90.2<\/td>92.2<\/strong><\/td><\/tr>
 <\/td>IF-Eval\uff08\u63d0\u793a\u4e25\u683c\uff09<\/td>86.5<\/strong><\/td>84.3<\/td>86.1<\/td>84.8<\/td>\u2013<\/td>83.3<\/td><\/tr>
 <\/td>GPQA-\u94bb\u77f3\u7ea7 (Pass@1)<\/td>65.0<\/td>49.9<\/td>59.1<\/td>60.0<\/td>75.7<\/strong><\/td>71.5<\/td><\/tr>
 <\/td>SimpleQA\uff08\u6b63\u786e\uff09<\/td>28.4<\/td>38.2<\/td>24.9<\/td>7.0<\/td>47.0<\/strong><\/td>30.1<\/td><\/tr>
 <\/td>\u6846\u67b6\uff08\u914d\u4ef6\uff09<\/td>72.5<\/td>80.5<\/td>73.3<\/td>76.9<\/td>\u2013<\/td>82.5<\/strong><\/td><\/tr>
 <\/td>AlpacaEval2.0 (LC-\u80dc\u7387)<\/td>52.0<\/td>51.1<\/td>70.0<\/td>57.8<\/td>\u2013<\/td>87.6<\/strong><\/td><\/tr>
 <\/td>ArenaHard\uff08GPT-4-1106\uff09<\/td>85.2<\/td>80.4<\/td>85.5<\/td>92.0<\/td>\u2013<\/td>92.3<\/strong><\/td><\/tr>
\u4ee3\u7801<\/td>LiveCodeBench (Pass@1-COT)<\/td>33.8<\/td>34.2<\/td>\u2013<\/td>53.8<\/td>63.4<\/td>65.9<\/strong><\/td><\/tr>
 <\/td>Codeforces\uff08\u767e\u5206\u4f4d\u6570\uff09<\/td>20.3<\/td>23.6<\/td>58.7<\/td>93.4<\/td>96.6<\/strong><\/td>96.3<\/td><\/tr>
 <\/td>Codeforces\uff08\u8bc4\u7ea7\uff09<\/td>717<\/td>759<\/td>1134<\/td>1820<\/td>2061<\/strong><\/td>2029<\/td><\/tr>
 <\/td>SWE \u5df2\u9a8c\u8bc1\uff08\u5df2\u89e3\u51b3\uff09<\/td>50.8<\/strong><\/td>38.8<\/td>42.0<\/td>41.6<\/td>48.9<\/td>49.2<\/td><\/tr>
 <\/td>Aider-Polyglot (Acc.)<\/td>45.3<\/td>16.0<\/td>49.6<\/td>32.9<\/td>61.7<\/strong><\/td>53.3<\/td><\/tr>
\u6570\u5b66<\/td>AIME 2024\uff08\u901a\u884c\u8bc1@1\uff09<\/td>16.0<\/td>9.3<\/td>39.2<\/td>63.6<\/td>79.2<\/td>79.8<\/strong><\/td><\/tr>
 <\/td>\u6570\u5b66-500 (\u901a\u8fc7@1)<\/td>78.3<\/td>74.6<\/td>90.2<\/td>90.0<\/td>96.4<\/td>97.3<\/strong><\/td><\/tr>
 <\/td>CNMO 2024 (\u901a\u884c\u8bc1@1)<\/td>13.1<\/td>10.8<\/td>43.2<\/td>67.6<\/td>\u2013<\/td>78.8<\/strong><\/td><\/tr>
\u4e2d\u6587<\/td>CLUEWSC\uff08EM\uff09<\/td>85.4<\/td>87.9<\/td>90.9<\/td>89.9<\/td>\u2013<\/td>92.8<\/strong><\/td><\/tr>
 <\/td>C-\u8bc4\u4f30\uff08EM\uff09<\/td>76.7<\/td>76.0<\/td>86.5<\/td>68.9<\/td>\u2013<\/td>91.8<\/strong><\/td><\/tr>
 <\/td>C-SimpleQA\uff08\u6b63\u786e\uff09<\/td>55.4<\/td>58.7<\/td>68.0<\/strong><\/td>40.3<\/td>\u2013<\/td>63.7<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n
\u84b8\u998f\u6a21\u578b\u8bc4\u4f30<\/h3>\n\n\n\n
\u6a21\u578b<\/th>AIME 2024 \u901a\u884c\u8bc1@1<\/th>AIME 2024 \u7f3a\u70b9@64<\/th>MATH-500 \u901a\u8fc7@1<\/th>GPQA \u94bb\u77f3\u901a\u884c\u8bc1@1<\/th>LiveCodeBench \u901a\u884c\u8bc1@1<\/th>CodeForces \u8bc4\u7ea7<\/th><\/tr><\/thead>
GPT-4o-0513<\/td>9.3<\/td>13.4<\/td>74.6<\/td>49.9<\/td>32.9<\/td>759<\/td><\/tr>
\u514b\u52b3\u5fb7-3.5-\u5341\u56db\u884c\u8bd7-1022<\/td>16.0<\/td>26.7<\/td>78.3<\/td>65.0<\/td>38.9<\/td>717<\/td><\/tr>
o1-\u8ff7\u4f60<\/td>63.6<\/td>80.0<\/td>90.0<\/td>60.0<\/td>53.8<\/td>1820<\/strong><\/td><\/tr>
QwQ-32B-\u9884\u89c8<\/td>44.0<\/td>60.0<\/td>90.6<\/td>54.5<\/td>41.9<\/td>1316<\/td><\/tr>
DeepSeek-R1-Distill-Qwen-1.5B<\/td>28.9<\/td>52.7<\/td>83.9<\/td>33.8<\/td>16.9<\/td>954<\/td><\/tr>
DeepSeek-R1-Distill-Qwen-7B<\/td>55.5<\/td>83.3<\/td>92.8<\/td>49.1<\/td>37.6<\/td>1189<\/td><\/tr>
DeepSeek-R1-Distill-Qwen-14B<\/td>69.7<\/td>80.0<\/td>93.9<\/td>59.1<\/td>53.1<\/td>1481<\/td><\/tr>
DeepSeek-R1-Distill-Qwen-32B<\/td>72.6<\/strong><\/td>83.3<\/td>94.3<\/td>62.1<\/td>57.2<\/td>1691<\/td><\/tr>
DeepSeek-R1-Distill-Llama-8B<\/td>50.4<\/td>80.0<\/td>89.1<\/td>49.0<\/td>39.6<\/td>1205<\/td><\/tr>
DeepSeek-R1-Distill-Llama-70B<\/td>70.0<\/td>86.7<\/strong><\/td>94.5<\/strong><\/td>65.2<\/strong><\/td>57.5<\/strong><\/td>1633<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n
Facebook<\/a>Line<\/a>Telegram<\/a>X<\/a>Reddit<\/a>Pinterest<\/a>\u5206\u4eab<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"
\u6700\u8fd1\uff0c\u4e00\u5bb6\u540d\u53ebDeepSeek\u7684\u521d\u521b\u516c\u53f8\u7ecf\u8fc7\u6280\u672f\u8fed\u4ee3\u4e0e\u5347\u7ea7\uff0c\u53d1\u5e03\u4e86\u5168\u65b0\u4e00\u4ee3\u5927\u6a21\u578b\uff0c\u201cDeepSeek […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[],"class_list":["post-36","post","type-post","status-publish","format-standard","hentry","category-uncategorized"],"_links":{"self":[{"href":"https:\/\/jp5.ssqwert.eu.org\/index.php\/wp-json\/wp\/v2\/posts\/36","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/jp5.ssqwert.eu.org\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/jp5.ssqwert.eu.org\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/jp5.ssqwert.eu.org\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/jp5.ssqwert.eu.org\/index.php\/wp-json\/wp\/v2\/comments?post=36"}],"version-history":[{"count":1,"href":"https:\/\/jp5.ssqwert.eu.org\/index.php\/wp-json\/wp\/v2\/posts\/36\/revisions"}],"predecessor-version":[{"id":37,"href":"https:\/\/jp5.ssqwert.eu.org\/index.php\/wp-json\/wp\/v2\/posts\/36\/revisions\/37"}],"wp:attachment":[{"href":"https:\/\/jp5.ssqwert.eu.org\/index.php\/wp-json\/wp\/v2\/media?parent=36"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/jp5.ssqwert.eu.org\/index.php\/wp-json\/wp\/v2\/categories?post=36"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/jp5.ssqwert.eu.org\/index.php\/wp-json\/wp\/v2\/tags?post=36"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}