LMArena — AI 模型人类盲测偏好 Elo 排名

LMArena Elo 榜

人类盲测偏好投票:模型有多强(与 OpenRouter/OpenCode 的「用量」互补)

数据日期 2026-08-23
对话之王
claude-fable-5
Elo 1508
编程之王
claude-opus-5-max
Elo 1691
开源 vs 闭源
0 : 70
开源模型数 : 闭源模型数
总投票数
112.3万
覆盖 70 个模型

对话能力 Elo Top10

· 盲测 A/B 投票

编程能力 Elo Top10

· 编程题盲测

对话能力 Elo 完整榜

#模型Elo投票
1
claude-fable-5
Anthropic
15082.4万
2
claude-opus-4-6-high
Anthropic
15047.2万
3
claude-opus-4-7-high
Anthropic
15026.0万
4
muse-spark-1.2 (xHigh)
Meta
14983.3k
5
claude-opus-4-6
Anthropic
14977.6万
6
claude-opus-4-7
Anthropic
14946.1万
7
claude-opus-5-high
Anthropic
14932.8万
8
muse-spark-1.1
Meta
14912.0万
9
gemini-3.7-flash-high
Google
14905.7k
10
kimi-k3-max
Moonshot
14891.5万
11
muse-spark
Meta
14881.4万
12
claude-opus-5-max
Anthropic
14871.3万
13
glm-5.3-max
Z.ai
14873.8k
14
gemini-3.1-pro-preview
Google
14869.9万
15
gemini-3-pro
Google
14854.1万
16
gpt-5.6-sol-xhigh
OpenAI
14822.0万
17
gpt-5.5-high
OpenAI
14826.0万
18
claude-opus-4-8-high
Anthropic
14824.5万
19
qwen3.8-max
Alibaba
148110.0k
20
gemini-3.6-flash-high
Google
14811.8万

编程能力 Elo 完整榜

#模型Elo投票
1
claude-opus-5-max
Anthropic
16918.1k
2
kimi-k3-max
Moonshot
16744.5k
3
qwen3.8-max
Alibaba
16693.2k
4
claude-opus-5-high
Anthropic
16638.7k
5
grok-4.6-high
SpaceXAI
16291.5k
6
claude-fable-5
Anthropic
16268.4k
7
gpt-5.6-sol-xhigh (codex-harness)
OpenAI
16199.5k
8
glm-5.3-max
Z.ai
15991.9k
9
qwen3.8-27b
Alibaba
15952.5k
10
gemini-3.7-flash-high
Google
15872.6k
11
glm-5.2-max
Z.ai
15828.8k
12
deepseek-v4-pro-high-20260813
DeepSeek
15822.9k
13
deepseek-v4-flash-high
DeepSeek
15793.5k
14
claude-opus-4-8-high
Anthropic
15631.2万
15
claude-opus-4-7
Anthropic
15581.5万
16
claude-opus-4-7-high
Anthropic
15571.5万
17
grok-4.5
SpaceXAI
15566.3k
18
claude-opus-4-6-high
Anthropic
15461.7万
19
claude-opus-4-8
Anthropic
15391.1万
20
muse-spark-1.1
Meta
15396.2k
21
gemini-3.6-flash-high
Google
15396.5k
22
claude-sonnet-5-high
Anthropic
15397.3k
23
claude-opus-4-6
Anthropic
15361.8万
24
muse-spark-1.2 (xHigh)
Meta
15342.1k
25
claude-sonnet-4-6
Anthropic
15222.0万
26
seed-2.1-pro-preview
Bytedance
15217.7k
27
gpt-5.6-terra-xhigh (codex-harness)
OpenAI
15205.7k
28
hy3
Tencent
15182.7k
29
gpt-5.6-luna-xhigh (codex-harness)
OpenAI
15185.7k
30
qwen3.7-max-20260517
Alibaba
15178.4k
31
glm-5.1
Z.ai
15099.3k
32
kimi-k2.6
Moonshot
15099.6k
33
gpt-5.5-xhigh (codex-harness)
OpenAI
15081.4万
34
gemini-3.5-flash-high
Google
14997.6k
35
claude-opus-4-5-20251101-high-32k
Anthropic
14941.1万
36
gemini-3.5-flash-medium
Google
14901.0万
37
minimax-m3
MiniMax
14881.1万
38
gpt-5.5-high (codex-harness)
OpenAI
14861.6万
39
qwen3.6-max-preview
Alibaba
14792.6k
40
mimo-v2.5-pro
Xiaomi
14761.4万
41
kimi-k2.7-code
Moonshot
14734.7k
42
claude-opus-4-5-20251101
Anthropic
14681.3万
43
deepseek-v4-pro-high-preview
DeepSeek
14641.2万
44
gpt-5.4-high (codex-harness)
OpenAI
14631.3k
45
qwen3.6-plus
Alibaba
14601.6万
46
gpt-5.5 (codex-harness)
OpenAI
14581.4万
47
gemini-3.5-flash-lite
Google
1449228
48
gemini-3.1-pro-preview
Google
14462.2万
49
deepseek-v4-pro
DeepSeek
14451.3万
50
gpt-5.4-medium (codex-harness)
OpenAI
14421.3k
数据来源: LMArena (Chatbot Arena) · 680万+ 盲测投票 · 经社区每日归档,6h 同步