三个皮匠报告—— 严选型全行业研究报告分享下载平台,您的专属行业智库!
欢迎来到三个皮匠报告!
登录
帮助中心
首页
最新
研搜
BETA
合集
英文
会议
数据
财报库
政策库
自研报告
入驻
您的当前位置:
首页
>
行业数据
>
不同规模模型在SWE-Bench验证常规性能表现对比
不同规模模型在SWE-Bench验证常规性能表现对比
行业数据
下载Excel
下载图片
原图定位
图片属性
图片格式:PNG
图片大小:186KB
图片尺寸:1838*1002
同报告图片
/
4
0
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
所属报告:
【华鑫证券】计算机行业周报:MistraiAI发布Devstral2系列,GPT-5.2定义专家级智能-251216(20页).pdf
打包全文图表
相关数据
SWE-benchVerified2026Q1实测得分
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
收藏
行业数据
2026-04-20
原图定位
查看详情
Qwen3Coder在SWE-Bench排名第四(截至2025/8/7)
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
收藏
行业数据
2025-08-13
原图定位
查看详情
每日贡献行数测试中,SWE-1与Claude系
fCDluo/lj7cgfCDmqKHlnovlkI3np7AgICAgICAgfCDmr4/ml6XotKHnjK7ooYzmlbAgfAp8LS0tLS0tfC0tLS0tLS0tLS0tLS0tLS18LS0tLS0tLS0tLS0tLS18CnwgMSAgICB8IENsYXVkZSBTb21tZXQgMy43IHwgMTUwMOW3puWPsyAgICAgfAp8IDIgICAgfCBTV0UtMSAgICAgICAgICB8IDEyMDDlt6blj7MgICAgIHwKfCAzICAgIHwgU1dFLTEgTGl0ZSAgICAgfCA5MDDlt6blj7MgICAgICB8CnwgNCAgICB8IENsYXVkZSBTb21tZXQgMy41IHwgNzAw5bem5Y+zICAgICAgfAp8IDUgICAgfCBDbGF1ZGUgSGFpa3UgMy41IHwgNjAw5bem5Y+zICAgICAgfAoK6LWE5paZ5p2l5rqQ77yaV2luZHN1cmblrpjnvZHvvIzkuJzljJfor4HliLg=
收藏
行业数据
2025-07-14
原图定位
查看详情
贡献率测试中,SWE-1与Claude系列的比
fCDluo/lj7cgfCDmqKHlnosgICAgICAgfCDotKHnjK7njocgICB8CnwtLS0tLS18LS0tLS0tLS0tLS0tfC0tLS0tLS0tLS18CnwgMSAgICB8IFNXRS0xICAgICAgfCA2MCUgICAgICB8CnwgMiAgICB8IENsYXVkZSBTb21tZXQgMy43IHwgNTAlICAgICAgfAp8IDMgICAgfCBDbGF1ZGUgU29tbWV0IDMuNSB8IDUwJSAgICAgIHwKfCA0ICAgIHwgU1dFLTEgbGl0ZSB8IDUwJSAgICAgIHwKfCA1ICAgIHwgQ2xhdWRlIEhha2t1IDMuNSB8IDQwJSAgICAgIHwKCgrotYTmlpnmnaXmupDvvJpXaW5kc3VyZuWumOe9kSwg5Lic5YyX6K+B5Yi4
收藏
行业数据
2025-07-14
原图定位
查看详情
在BIG-Bench任务中,>92%的涌现能力是这两个度量下出现的
5qC55o2u5oKo5o+Q5L6b55qE5L+h5oGv77yM5Lul5LiL5piv5pWw5o2u6KGo5qC85ZKM6LWE5paZ5p2l5rqQ77yaCgp8IOWkmumhuemAieaLqeivhOWIhiB8IOadoeS7tiB8CnwtLS0tLS0tLS0tLS0tLXwtLS0tLS18CnwgMSAgICAgICAgICAgIHwg5aaC5p6c5pyA6auY5qaC546H5Zyo5q2j56Gu6YCJ6aG55LiKIHwKfCAwICAgICAgICAgICAgfCDlkKbliJkgfAoKfCDnsr7noa7lrZfnrKbkuLLljLnphY0gfCDmnaHku7YgfAp8LS0tLS0tLS0tLS0tLS0tLXwtLS0tLS18CnwgMSAgICAgICAgICAgICAgfCDlpoLmnpzovpPlh7rlrZfnrKbkuLLkuI7nm67moIflrZfnrKbkuLLlrozlhajljLnphY0gfAp8IDAgICAgICAgICAgICAgIHwg5ZCm5YiZIHwKCui1hOaWmeadpea6kO+8muOAikFyZSBFbWVyZ2VudCBBYmlsaXRpZXMgb2YgTGFyZ2UgTGFuZ3VhZ2UgTW9kZWxzIGEgTWlyYWdlP+OAi++8jOS4reS/oeW7uuaKlQ==
收藏
行业数据
2023-05-15
原图定位
查看详情
最新数据
全球全年龄段奶粉市场规模(亿美元)
收藏
行业数据
2026-07-30
原图定位
查看详情
全球酸奶及发酵乳制品市场规模(亿美元)
收藏
行业数据
2026-07-30
原图定位
查看详情
全球药用级益生菌市场规模(百万美元)
收藏
行业数据
2026-07-30
原图定位
查看详情
全球益生菌膳食补充剂市场规模及增速
收藏
行业数据
2026-07-30
原图定位
查看详情
全球拥有AKK专利菌株的核心企业
收藏
行业数据
2026-07-30
原图定位
查看详情
Mistral官方指出,Devstral2(123B)和DevstralSmall2(24B)分别比DeepSeekV3.2小 5 倍和 28 倍,比 KimiK2 小 8 倍和 41 倍。这证明了紧凑型模型可以匹敌甚至超越更大型竞争对手的性能。
行业数据
原图定位
当前图表所属:
【华鑫证券】计算机行业周报:MistraiAI发布Devstral2系列,GPT-5.2定义专家级智能-251216(20页).pdf
GPT‑5.2Thinking在GDPval上的表现
GPT-5.2在多项基准测试中的表现
修改版MIT许可证编码
2025.12.05-2025.12.11AI相关网站流量
下载图片或Excel,需升级至尊VIP会员或「预见者」VIP
立即去升级
客服
商务合作
小程序
服务号
回到顶部
折叠
微信扫码登录
微信扫码登录
双重福利,扫码领取!
关注公众号『三个皮匠报告』,免费领取报告
1.100+最新优质行业报告
2.每周免费精选报告分享
手机快捷登录
账号登录
记住密码
忘记密码?
验证即登录,未注册将自动创建账号
获取验证码
收不到验证码注意事项
登录代表您已阅读并同意
《三个皮匠服务条款》