1
G
gemini-3-pro
gemini-3-proGoogle
100.0
224
1.05M
¥14.4 / ¥86.4Input/Output
Chat · Vision · Captioning Leaderboard
Ranking for Vision / Captioning, based on public preference data.
Captioning model ranking guide
Ranking for Vision / Captioning, based on public preference data.
gemini-3-progemini-3.1-pro-previewgemini-2.5-progemini-3-flashgpt-5.2-high
2
Ggemini-3.1-pro-preview
gemini-3.1-pro-previewGoogle
96.4
86
1.05M
¥14.4 / ¥86.4Input/Output
3
Ggemini-2.5-pro
gemini-2.5-proGoogle
92.9
808
1.05M
¥9 / ¥72Input/Output
4
Ggemini-3-flash
gemini-3-flashGoogle
89.3
157
1.05M
¥3.6 / ¥21.6Input/Output
5
O
gpt-5.2-high
gpt-5.2-highOpenai
85.7
79
400K
¥12.6 / ¥101Input/Output
6
A
qwen3-vl-235b-a22b-instruct
qwen3-vl-235b-a22b-instructAlibaba
82.1
148
128K
¥2.16 / ¥8.64Input/Output
7
Ggemini-2.5-flash
gemini-2.5-flashGoogle
78.6
596
1.05M
¥2.16 / ¥18Input/Output
8
M
kimi-k2.5-thinking
kimi-k2.5-thinkingMoonshot
75.0
89
262K
¥4.32 / ¥21.6Input/Output
9
Ogpt-5.1-high
gpt-5.1-highOpenai
71.4
100
400K
¥9 / ¥72Input/Output
10
X
grok-4-0709
grok-4-0709Xai
67.9
377
256K
¥21.6 / ¥108Input/Output
11
Ggemini-3-flash (thinking-minimal)
gemini-3-flash (thinking-minimal)Google
64.3
138
1.05M
¥3.6 / ¥21.6Input/Output
12
Ggemma-4-31b
gemma-4-31bGoogle
60.7
50
262K
¥3.24 / ¥7.2Input/Output
13
Ochatgpt-4o-latest-20250326
chatgpt-4o-latest-20250326Openai
57.1
286
128K
¥18 / ¥72Input/Output
14
Ogpt-5-chat
gpt-5-chatOpenai
53.6
403
400K
¥9 / ¥72Input/Output
15
Ogpt-5-mini-high
gpt-5-mini-highOpenai
50.0
302
400K
¥1.8 / ¥14.4Input/Output
16
Oo3-2025-04-16
o3-2025-04-16Openai
46.4
564
200K
¥14.4 / ¥57.6Input/Output
17
Ogpt-4.1-2025-04-14
gpt-4.1-2025-04-14Openai
42.9
443
1.05M
¥14.4 / ¥57.6Input/Output
18
Ogpt-5-high
gpt-5-highOpenai
39.3
382
400K
¥9 / ¥72Input/Output
19
Oo4-mini-2025-04-16
o4-mini-2025-04-16Openai
35.7
442
200K
¥7.92 / ¥31.7Input/Output
20
Ogpt-5.1
gpt-5.1Openai
32.1
125
400K
¥9 / ¥72Input/Output
21
Ggemini-2.5-flash-lite-preview-06-17-thinking
gemini-2.5-flash-lite-preview-06-17-thinkingGoogle
28.6
406
65.5K
¥0.72 / ¥2.88Input/Output
22
Ogpt-4.1-mini-2025-04-14
gpt-4.1-mini-2025-04-14Openai
25.0
410
1.05M
¥2.88 / ¥11.5Input/Output
23
Ogpt-5.2
gpt-5.2Openai
21.4
87
400K
¥12.6 / ¥101Input/Output
24
MA
mistral-medium-2508
mistral-medium-2508Mistral
17.9
412
262K
¥2.88 / ¥14.4Input/Output
25
MAmistral-small-3.1-24b-instruct-2503
mistral-small-3.1-24b-instruct-2503Mistral
14.3
281
262K
¥2.88 / ¥14.4Input/Output
26
Ggemma-3-27b-it
gemma-3-27b-itGoogle
10.7
273
128K
¥2.15 / ¥2.15Input/Output
27
Ggemini-2.0-flash-001
gemini-2.0-flash-001Google
7.1
110
1.05M
¥1.08 / ¥4.32Input/Output
28
MAmistral-medium-2505
mistral-medium-2505Mistral
3.6
168
262K
¥2.88 / ¥14.4Input/Output
29
MAmistral-small-2506
mistral-small-2506Mistral
0.0
194
262K
¥2.88 / ¥14.4Input/Output
gemini-3-pro why it ranks first
gemini-3-pro ranks first with a percent score of 100.0 and 224 samples. Use it as the first option for this leaderboard, then compare price, context and availability.
Do not only look at rank #1
Start with the leaderboard closest to your task. Compare the top models by score and sample size, then check price, context length, open or closed access, and provider availability.
Compare adjacent capabilities
FAQ
图像描述排行榜看什么指标？
主要看排名、百分制分数、样本量和来源。分数用于快速比较同一榜单内模型表现，样本量用于判断结果稳定性。
为什么不同榜单不能直接混合成总分？
不同榜单的任务、样本和评测口径不同，模力榜默认只在同一榜单内排序，避免把写作、代码、图像等能力强行合并。
图像描述模型应该怎么选？
优先看与你任务最接近的榜单，再结合价格、上下文长度、开源闭源和厂商可用性。排名靠前不代表适合所有预算和部署方式。
榜单多久更新？
页面展示的是最新成功采集的公开榜单数据。当前优先使用 LMArena leaderboard dataset，并在页面来源中保留原始链接。