跳到主内容
知行录
leaderboard.cn
Toggle Sidebar
搜索
⌘K
分析
前沿模型
历史趋势
本地模型
资料库
模型库
厂商
基准评测
数据源
显示模式
Toggle Sidebar
知行录
leaderboard.cn
正在加载页面内容
正在加载页面内容
首页
长上下文
检索、归纳与百万级上下文推理。
更新于 07/16 02:55
模型
76 个 · 16 项基准
评分
Graphwalks BFS 1M F1
Graphwalks BFS 256k F1
Arena Document Rating
OpenAI MRCR v2 8-needle 512K-1M
OpenAI MRCR v2 8-needle 256K-512K
CL-bench
AA-LCR
CL-bench life
Graphwalks parents 1M F1
MRCR 1M
OpenAI MRCR v2 8-needle 128K-256K
Graphwalks parents 256k F1
OpenAI MRCR v2 8-needle 64K-128K
OpenAI MRCR v2 8-needle 16K-32K
OpenAI MRCR v2 8-needle 32K-64K
OpenAI MRCR v2 8-needle 4K-8K
1
GPT-5.6 Sol
100.0
77.1%
90.7%
-
73.8%
91.5%
-
-
-
-
-
-
-
-
-
-
-
2
Claude Fable 5
97.2
79.4%
91.1%
1507.1 Elo
-
-
-
-
-
-
-
-
-
-
-
-
-
3
GPT-5.6 Terra
92.6
71.2%
76.9%
-
72.5%
89.6%
-
-
-
-
-
-
-
-
-
-
-
4
Claude Opus 4.6
91.5
41.2%
-
1508.1 Elo
-
-
-
-
-
72%
92.9%
-
-
-
-
-
-
5
GPT-5.5
89.2
45.4%
73.7%
1488.2 Elo
74%
81.5%
27.8%
76.4%
21.1%
58.5%
74%
87.5%
90.1%
83.1%
96.5%
90%
98.1%
6
Claude Sonnet 4.6
88.8
-
-
1487.6 Elo
-
-
-
-
-
-
-
-
-
-
-
-
-
7
GPT-5.6 Luna
88.0
51.2%
81.3%
-
41.3%
41.3%
-
-
-
-
-
-
-
-
-
-
-
8
Claude Opus 4.8
87.9
68.1%
85.9%
1476.6 Elo
-
-
24.8%
72.2%
16.9%
-
-
-
-
-
-
-
-
9
Gemini 3.5 Flash
87.8
-
-
1481.0 Elo
-
-
-
-
-
-
-
-
-
-
-
-
-
10
混元3
86.2
-
-
-
-
-
23.8%
73.4%
17%
-
-
-
-
-
-
-
-
11
Claude Opus 4.7
86.0
-
76.9%
1504.5 Elo
32.2%
-
22.9%
-
-
-
-
59.2%
93.6%
-
-
-
-
12
Claude Sonnet 5
85.8
-
-
1476.1 Elo
-
-
-
-
-
-
-
-
-
-
-
-
-
13
GLM-5.2
85.3
-
-
-
-
-
23.1%
73.4%
21%
-
-
-
-
-
-
-
-
14
Claude Opus 4.5
84.7
-
-
-
-
-
-
74%
-
-
-
-
-
-
-
-
-
15
GPT-5.4
83.6
9.4%
62.5%
1474.9 Elo
36.6%
57.5%
-
-
-
44.4%
-
79.3%
82.8%
86%
97.2%
90.5%
97.3%
16
Muse Spark
83.0
-
-
1444.9 Elo
-
-
-
-
-
-
-
-
-
-
-
-
-
17
Gemini 3.1 Pro
82.9
-
-
1441.4 Elo
-
-
21.3%
74.6%
16.5%
-
76.3%
-
-
-
-
-
-
18
Qwen3.7-Plus
82.1
-
-
1444.2 Elo
-
-
-
-
-
-
-
-
-
-
-
-
-
19
GPT-5.2
81.9
-
-
1405.4 Elo
-
-
-
-
-
-
-
77%
-
85.6%
95.3%
92%
98.2%
20
Seed2.1 Turbo
81.6
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
21
Kimi-K2.6
81.4
-
-
1453.6 Elo
-
-
-
70.2%
-
-
-
-
-
-
-
-
-
22
DeepSeek-V4-Pro
80.5
-
-
-
-
-
18.1%
71.3%
14.3%
-
83.5%
-
-
-
-
-
-
23
Seed2.1 Pro
80.4
-
-
-
-
-
23.5%
70%
13.3%
-
-
-
-
-
-
-
-
24
GPT-5
80.1
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
低
高
单一有效来源基准按 0.5 权重计入,Arena.ai 除外;最高模型归一化为 100