模型排行榜
每日同步
来源日期:2026-11-19
内容更新:2026-11-20 18:04(UTC+8)
视觉理解
视觉理解综合榜,采用官方默认的风格校正评分;竞技场评分越高越好。
| 排名 | 模型 | 机构 | 竞技场评分 | 置信区间 | 投票数 |
|---|---|---|---|---|---|
| 1 | Claude Fable 5 (High) | Anthropic | 1,308.8 | 1,301.9 – 1,315.6 | 15,449 |
| 2 | Claude Fable 5.2 (Max) | Anthropic | 1,305.3 | 1,287.1 – 1,323.5 | 1,176 |
| 3 | Claude Opus 4.7 | Anthropic | 1,298.7 | 1,292.3 – 1,305.0 | 24,930 |
| 4 | Claude Opus 4.6 (High) | Anthropic | 1,298.2 | 1,291.9 – 1,304.6 | 22,067 |
| 5 | Claude Opus 4.7 (High) | Anthropic | 1,298.0 | 1,291.5 – 1,304.5 | 22,530 |
| 6 | Gemini 3.7 Flash (High) | 谷歌 | 1,294.2 | 1,285.7 – 1,302.7 | 6,776 |
| 7 | Muse Spark | Meta | 1,294.0 | 1,285.0 – 1,303.1 | 5,835 |
| 8 | Qwen3.8-Max | 阿里巴巴 | 1,293.7 | 1,286.4 – 1,301.0 | 10,833 |
| 9 | Claude Opus 5.5 (High) | Anthropic | 1,293.2 | 1,284.0 – 1,302.4 | 5,903 |
| 10 | Muse Spark 1.2 (xHigh) | Meta | 1,293.0 | 1,278.6 – 1,307.4 | 1,922 |
| 11 | Claude Opus 4.6 | Anthropic | 1,293.0 | 1,286.8 – 1,299.2 | 26,552 |
| 12 | Gemini 3.8 Flash (High) | 谷歌 | 1,292.8 | 1,285.2 – 1,300.4 | 9,782 |
| 13 | GPT-6.1 Sol (Max) | OpenAI | 1,289.9 | 1,280.2 – 1,299.6 | 5,043 |
| 14 | Claude Fable 5.1 (Max) | Anthropic | 1,289.8 | 1,281.6 – 1,298.0 | 7,185 |
| 15 | Gemini 3 Pro | 谷歌 | 1,289.1 | 1,281.7 – 1,296.5 | 13,482 |
| 16 | Muse Spark 1.3 (Max) | Meta | 1,288.2 | 1,280.7 – 1,295.8 | 9,287 |
| 17 | GPT-5.6 Sol (xHigh) | OpenAI | 1,288.1 | 1,281.0 – 1,295.3 | 12,244 |
| 18 | Muse Spark 1.3 (xHigh) | Meta | 1,287.8 | 1,280.3 – 1,295.4 | 9,809 |
| 19 | Claude Opus 4.8 (High) | Anthropic | 1,285.5 | 1,279.0 – 1,291.9 | 18,113 |
| 20 | Claude Opus 5 (High) | Anthropic | 1,285.3 | 1,278.8 – 1,291.7 | 19,229 |
| 21 | GPT-5.5 | OpenAI | 1,284.8 | 1,279.1 – 1,290.4 | 29,925 |
| 22 | GPT-5.5 (High) | OpenAI | 1,283.9 | 1,278.2 – 1,289.6 | 28,221 |
| 23 | Gemini 3.5 Flash (Medium) | 谷歌 | 1,282.1 | 1,275.5 – 1,288.7 | 15,320 |
| 24 | GPT-5.4 (High) | OpenAI | 1,282.0 | 1,276.4 – 1,287.6 | 31,907 |
| 25 | Gemini 3.5 Flash (High) | 谷歌 | 1,281.1 | 1,274.5 – 1,287.7 | 15,228 |
| 26 | GPT-6 Astra (Max) | OpenAI | 1,281.0 | 1,272.3 – 1,289.7 | 6,436 |
| 27 | Gemini 3.1 Pro Preview | 谷歌 | 1,280.4 | 1,275.4 – 1,285.4 | 48,038 |
| 28 | Grok 4.5 | xai | 1,280.3 | 1,273.4 – 1,287.1 | 13,903 |
| 29 | Gemini 3.6 Flash (High) | 谷歌 | 1,280.0 | 1,272.5 – 1,287.5 | 10,275 |
| 30 | Muse Spark 1.1 | Meta | 1,278.9 | 1,272.1 – 1,285.7 | 13,311 |
| 31 | Claude Opus 4.8 | Anthropic | 1,278.1 | 1,271.6 – 1,284.5 | 18,588 |
| 32 | GPT-5.5 Instant | OpenAI | 1,277.7 | 1,269.0 – 1,286.4 | 7,549 |
| 33 | GPT-5.2 Chat (20260210) | OpenAI | 1,277.5 | 1,271.1 – 1,284.0 | 16,336 |
| 34 | GPT-5.4 | OpenAI | 1,277.3 | 1,270.9 – 1,283.7 | 22,546 |
| 35 | Claude Sonnet 4.6 | Anthropic | 1,274.7 | 1,268.7 – 1,280.8 | 27,131 |
| 36 | GLM-5.3-Flash | zai | 1,273.8 | 1,268.4 – 1,279.2 | 36,497 |
| 37 | Gemini 3 Flash | 谷歌 | 1,272.2 | 1,267.1 – 1,277.3 | 38,733 |
| 38 | GPT-5.6 Terra (xHigh) | OpenAI | 1,267.1 | 1,260.1 – 1,274.1 | 13,166 |
| 39 | Gemini 3.5 Flash-Lite | 谷歌 | 1,265.9 | 1,258.4 – 1,273.4 | 10,713 |
| 40 | Kimi K2.6 | 月之暗面 | 1,264.7 | 1,257.9 – 1,271.4 | 16,241 |
| 41 | Gemma 4 31B | 谷歌 | 1,264.7 | 1,259.2 – 1,270.1 | 42,217 |
| 42 | Grok 4.6 (High) | xai | 1,264.3 | 1,256.4 – 1,272.2 | 8,578 |
| 43 | Claude Sonnet 5.5 (xHigh) | Anthropic | 1,264.2 | 1,254.8 – 1,273.6 | 5,388 |
| 44 | GPT-6.1 Terra (Max) | OpenAI | 1,263.0 | 1,250.1 – 1,276.0 | 2,796 |
| 45 | Qwen3.7-Plus | 阿里巴巴 | 1,262.4 | 1,255.9 – 1,268.9 | 15,084 |
| 46 | DeepSeek-V4.1-Flash-20260910 (Max) | 深度求索 | 1,262.2 | 1,252.0 – 1,272.4 | 4,192 |
| 47 | Claude Sonnet 5 (High) | Anthropic | 1,260.8 | 1,254.1 – 1,267.5 | 14,701 |
| 48 | Gemini 3 Flash (Minimal) | 谷歌 | 1,259.8 | 1,254.6 – 1,265.1 | 36,810 |
| 49 | GPT-6 Sol (Max) | OpenAI | 1,257.5 | 1,248.6 – 1,266.5 | 6,202 |
| 50 | Dola Seed 2.0 Pro | 字节跳动 | 1,256.4 | 1,249.2 – 1,263.6 | 10,914 |
| 51 | GPT-5.6 Luna (xHigh) | OpenAI | 1,255.7 | 1,248.4 – 1,263.1 | 10,996 |
| 52 | Grok 4.20 Beta (Reasoning, 0309) | xai | 1,255.4 | 1,249.4 – 1,261.4 | 26,501 |
| 53 | Kimi K2.5 (Thinking) | 月之暗面 | 1,252.6 | 1,247.0 – 1,258.2 | 27,947 |
| 54 | GPT-5.1 (High) | OpenAI | 1,250.9 | 1,243.1 – 1,258.7 | 9,775 |
| 55 | Grok 4.20 Multi-Agent Beta (0309) | xai | 1,250.7 | 1,244.7 – 1,256.8 | 24,292 |
| 56 | Grok 4.7 (xHigh) | xai | 1,250.0 | 1,241.8 – 1,258.2 | 7,761 |
| 57 | GPT-5.4 mini (High) | OpenAI | 1,249.8 | 1,243.7 – 1,255.9 | 25,461 |
| 58 | MiMo-V2.6-Pro | 小米 | 1,247.5 | 1,232.8 – 1,262.1 | 1,812 |
| 59 | Gemini 2.5 Pro | 谷歌 | 1,247.5 | 1,243.0 – 1,252.0 | 88,464 |
| 60 | Step 5 Preview | 阶跃星辰 | 1,246.5 | 1,236.4 – 1,256.6 | 4,395 |
| 61 | MiMo-V2.6-Flash | 小米 | 1,245.8 | 1,238.1 – 1,253.6 | 9,656 |
| 62 | Qwen3.5-397B-A17B | 阿里巴巴 | 1,245.4 | 1,240.3 – 1,250.6 | 33,471 |
| 63 | Qwen3.8-27B | 阿里巴巴 | 1,244.4 | 1,237.2 – 1,251.5 | 10,910 |
| 64 | GPT-5.2 (High) | OpenAI | 1,242.7 | 1,236.7 – 1,248.7 | 18,881 |
| 65 | Gemma 4 26B A4B | 谷歌 | 1,241.2 | 1,234.8 – 1,247.5 | 23,681 |
| 66 | Grok 4.3 | xai | 1,241.1 | 1,235.5 – 1,246.8 | 29,102 |
| 67 | GPT-5.1 | OpenAI | 1,240.3 | 1,232.7 – 1,247.8 | 10,727 |
| 68 | ChatGPT-4o Latest (20250326) | OpenAI | 1,238.1 | 1,232.6 – 1,243.7 | 23,305 |
| 69 | Kimi K2.5 (Instant) | 月之暗面 | 1,235.5 | 1,224.8 – 1,246.2 | 4,085 |
| 70 | Gemini 3.1 Flash-Lite Preview | 谷歌 | 1,234.4 | 1,228.7 – 1,240.0 | 27,378 |
| 71 | GPT-6 Luna (Max) | OpenAI | 1,234.1 | 1,225.8 – 1,242.5 | 7,728 |
| 72 | MiMo-V2.5 | 小米 | 1,233.8 | 1,228.0 – 1,239.5 | 29,320 |
| 73 | Mistral Large 4 | mistral | 1,233.2 | 1,221.9 – 1,244.5 | 3,699 |
| 74 | MiniMax-M3 | MiniMax | 1,230.9 | 1,224.7 – 1,237.1 | 21,034 |
| 75 | Gemini 2.5 Flash Preview (09-2025) | 谷歌 | 1,229.2 | 1,219.3 – 1,239.0 | 4,759 |
| 76 | GPT-5.2 | OpenAI | 1,227.7 | 1,221.8 – 1,233.7 | 19,962 |
| 77 | Qwen3.5-122B-A10B | 阿里巴巴 | 1,225.6 | 1,218.9 – 1,232.3 | 13,158 |
| 78 | GPT-5 Chat | OpenAI | 1,225.5 | 1,218.4 – 1,232.6 | 39,816 |
| 79 | GPT-4.5 Preview (2025-02-27) | OpenAI | 1,225.4 | 1,213.7 – 1,237.2 | 2,925 |
| 80 | ERNIE-5.0-Preview (1220) | baidu | 1,218.7 | 1,207.5 – 1,229.9 | 3,330 |
| 81 | MiMo-V2-Omni | 小米 | 1,216.1 | 1,208.7 – 1,223.4 | 10,573 |
| 82 | o3 (2025-04-16) | OpenAI | 1,216.1 | 1,209.5 – 1,222.6 | 45,482 |
| 83 | Gemini 2.5 Flash | 谷歌 | 1,215.5 | 1,211.0 – 1,220.1 | 59,854 |
| 84 | Qwen3.5-27B | 阿里巴巴 | 1,215.3 | 1,209.9 – 1,220.6 | 31,190 |
| 85 | Qwen3-VL-235B-A22B-Instruct | 阿里巴巴 | 1,214.3 | 1,207.3 – 1,221.2 | 12,648 |
| 86 | GPT-4.1 (2025-04-14) | OpenAI | 1,213.9 | 1,207.2 – 1,220.6 | 40,925 |
| 87 | GPT-5 (High) | OpenAI | 1,211.5 | 1,204.4 – 1,218.5 | 34,338 |
| 88 | NVIDIA-Nemotron-3.5-Super-VL-120B-A12B (NVFP4) | nvidia | 1,211.1 | 1,202.5 – 1,219.8 | 7,022 |
| 89 | Inkling-Small | thinky | 1,208.1 | 1,200.2 – 1,215.9 | 9,182 |
| 90 | Claude Opus 4 (Thinking 16K, 20250514) | Anthropic | 1,207.9 | 1,191.8 – 1,224.0 | 1,288 |
| 91 | Claude Sonnet 4 (Thinking 32K, 20250514) | Anthropic | 1,206.8 | 1,189.8 – 1,223.7 | 1,142 |
| 92 | o4-mini (2025-04-16) | OpenAI | 1,201.3 | 1,194.2 – 1,208.5 | 41,255 |
| 93 | GPT-4.1 mini (2025-04-14) | OpenAI | 1,201.0 | 1,193.5 – 1,208.6 | 40,139 |
| 94 | Mistral Large 3 | mistral | 1,200.4 | 1,192.8 – 1,208.0 | 10,071 |
| 95 | Mistral Medium 3.5 | mistral | 1,198.9 | 1,189.3 – 1,208.4 | 5,327 |
| 96 | GPT-5.4 nano (High) | OpenAI | 1,198.6 | 1,192.3 – 1,204.8 | 25,347 |
| 97 | Grok 4.1 Fast (Reasoning) | xai | 1,196.6 | 1,189.7 – 1,203.4 | 13,282 |
| 98 | Claude 3.7 Sonnet (Thinking 32K, 20250219) | Anthropic | 1,193.1 | 1,177.6 – 1,208.7 | 1,431 |
| 99 | o1 (2024-12-17) | OpenAI | 1,193.0 | 1,182.2 – 1,203.8 | 3,694 |
| 100 | Claude Opus 4 (20250514) | Anthropic | 1,189.2 | 1,176.3 – 1,202.1 | 2,166 |
| 101 | Gemini 2.5 Flash-Lite Preview (Thinking, 06-17) | 谷歌 | 1,189.0 | 1,181.5 – 1,196.5 | 35,853 |
| 102 | Claude Sonnet 4 (20250514) | Anthropic | 1,189.0 | 1,174.9 – 1,203.0 | 1,802 |
| 103 | Qwen3-VL-235B-A22B-Thinking | 阿里巴巴 | 1,188.1 | 1,175.7 – 1,200.5 | 2,378 |
| 104 | GPT-5 mini (High) | OpenAI | 1,185.6 | 1,177.5 – 1,193.7 | 28,878 |
| 105 | Qwen-VL-Max (2025-08-13) | 阿里巴巴 | 1,184.3 | 1,172.2 – 1,196.4 | 3,157 |
| 106 | Grok 4 (0709) | xai | 1,184.0 | 1,176.7 – 1,191.4 | 30,106 |
| 107 | Gemini 1.5 Pro (002) | 谷歌 | 1,179.9 | 1,170.7 – 1,189.0 | 8,902 |
| 108 | Claude 3.7 Sonnet (20250219) | Anthropic | 1,175.6 | 1,166.2 – 1,185.0 | 4,515 |
| 109 | Gemini 2.5 Flash-Lite Preview (Non-Thinking, 09-2025) | 谷歌 | 1,173.8 | 1,164.2 – 1,183.4 | 4,841 |
| 110 | Gemini 2.0 Flash (001) | 谷歌 | 1,172.4 | 1,165.1 – 1,179.6 | 10,120 |
| 111 | GPT-4o (2024-05-13) | OpenAI | 1,162.1 | 1,153.3 – 1,171.0 | 23,273 |
| 112 | Claude 3.5 Sonnet (20241022) | Anthropic | 1,161.4 | 1,153.4 – 1,169.4 | 10,373 |
| 113 | GLM-4.6V | zai | 1,160.6 | 1,146.8 – 1,174.3 | 2,440 |
| 114 | Gemma 3 27B IT | 谷歌 | 1,160.1 | 1,152.0 – 1,168.2 | 16,928 |
| 115 | Mistral Medium 3.1 (2508) | mistral | 1,159.5 | 1,153.3 – 1,165.7 | 41,496 |
| 116 | Mistral Medium 3 (2505) | mistral | 1,155.2 | 1,146.9 – 1,163.5 | 10,769 |
| 117 | Step-1o Turbo (202506) | 阶跃星辰 | 1,154.9 | 1,140.5 – 1,169.2 | 1,894 |
| 118 | Hunyuan Vision 1.5 Thinking | 腾讯 | 1,153.8 | 1,141.5 – 1,166.0 | 2,414 |
| 119 | GLM-4.5V | zai | 1,152.8 | 1,141.1 – 1,164.6 | 3,236 |
| 120 | Hunyuan Large Vision | 腾讯 | 1,149.9 | 1,133.4 – 1,166.4 | 1,271 |
| 121 | GPT-5 nano (High) | OpenAI | 1,147.9 | 1,136.6 – 1,159.2 | 3,921 |
| 122 | Claude 3.5 Sonnet (20240620) | Anthropic | 1,146.0 | 1,135.8 – 1,156.1 | 21,624 |
| 123 | Llama 4 Maverick 17B 128E Instruct | Meta | 1,145.6 | 1,136.5 – 1,154.8 | 6,909 |
| 124 | Step-3 | 阶跃星辰 | 1,143.8 | 1,131.9 – 1,155.7 | 3,187 |
| 125 | Gemini 1.5 Flash (002) | 谷歌 | 1,142.0 | 1,132.0 – 1,152.0 | 7,241 |
| 126 | Mistral Small 3.2 (2506) | mistral | 1,140.1 | 1,131.0 – 1,149.2 | 10,814 |
| 127 | Gemini 2.0 Flash-Lite Preview (02-05) | 谷歌 | 1,136.1 | 1,125.6 – 1,146.6 | 3,991 |
| 128 | Llama 4 Scout 17B 16E Instruct | Meta | 1,127.8 | 1,118.2 – 1,137.3 | 6,433 |
| 129 | Claude 3.5 Haiku (20241022) | Anthropic | 1,127.6 | 1,111.5 – 1,143.6 | 1,397 |
| 130 | Mistral Small 3.1 24B Instruct (2503) | mistral | 1,127.0 | 1,118.5 – 1,135.5 | 28,440 |
| 131 | Step-1o Vision 32K (High Res) | 阶跃星辰 | 1,123.9 | 1,111.2 – 1,136.6 | 2,833 |
| 132 | Qwen2.5-VL-72B-Instruct | 阿里巴巴 | 1,121.3 | 1,110.4 – 1,132.2 | 3,768 |
| 133 | Gemini 1.5 Pro (001) | 谷歌 | 1,119.8 | 1,108.1 – 1,131.4 | 16,734 |
| 134 | GPT-4o (2024-08-06) | OpenAI | 1,118.4 | 1,106.0 – 1,130.7 | 3,376 |
| 135 | Qwen2.5-VL-32B-Instruct | 阿里巴巴 | 1,118.0 | 1,102.5 – 1,133.5 | 1,490 |
| 136 | Molmo 2 8B | allenai | 1,112.6 | 1,092.8 – 1,132.3 | 1,185 |
| 137 | GPT-4 Turbo (2024-04-09) | OpenAI | 1,112.4 | 1,100.8 – 1,124.0 | 13,391 |
| 138 | GPT-4o mini (2024-07-18) | OpenAI | 1,097.1 | 1,088.7 – 1,105.4 | 17,341 |
| 139 | Pixtral Large (2411) | mistral | 1,095.5 | 1,086.2 – 1,104.9 | 5,423 |
| 140 | GPT-4.1 nano (2025-04-14) | OpenAI | 1,088.9 | 1,070.6 – 1,107.2 | 1,211 |
| 141 | Qwen2-VL-72B | 阿里巴巴 | 1,085.1 | 1,074.8 – 1,095.3 | 5,937 |
| 142 | Qwen-VL-Max (1119) | 阿里巴巴 | 1,084.8 | 1,068.7 – 1,100.8 | 1,422 |
| 143 | Gemini 1.5 Flash-8B (001) | 谷歌 | 1,071.7 | 1,061.0 – 1,082.3 | 6,243 |
| 144 | Step-1V 32K | 阶跃星辰 | 1,064.4 | 1,047.6 – 1,081.2 | 1,534 |
| 145 | Claude 3 Opus (20240229) | Anthropic | 1,062.3 | 1,051.8 – 1,072.7 | 15,565 |
| 146 | Gemini 1.5 Flash (001) | 谷歌 | 1,059.6 | 1,047.7 – 1,071.5 | 13,260 |
| 147 | Molmo 72B (0924) | allenai | 1,045.6 | 1,032.4 – 1,058.7 | 3,048 |
| 148 | Hunyuan Standard Vision (2024-12-31) | 腾讯 | 1,042.1 | 1,021.1 – 1,063.1 | 809 |
| 149 | Llama 3.2 90B Vision Instruct | Meta | 1,032.7 | 1,023.4 – 1,042.0 | 8,682 |
| 150 | Qwen2-VL-7B-Instruct | 阿里巴巴 | 1,031.2 | 1,020.6 – 1,041.8 | 5,766 |
| 151 | Pixtral 12B (2409) | mistral | 1,025.4 | 1,015.8 – 1,035.0 | 7,511 |
| 152 | InternVL2 26B | 1,025.2 | 1,012.3 – 1,038.1 | 5,148 | |
| 153 | Amazon Nova Lite 1.0 | amazon | 1,019.3 | 1,004.3 – 1,034.4 | 1,854 |
| 154 | Amazon Nova Pro 1.0 | amazon | 1,019.0 | 1,005.1 – 1,032.8 | 2,335 |
| 155 | Claude 3 Sonnet (20240229) | Anthropic | 1,017.1 | 1,005.7 – 1,028.5 | 12,314 |
| 156 | Yi Vision | 1,003.0 | 984.9 – 1,021.1 | 1,219 | |
| 157 | Aya Vision 32B | cohere | 1,000.8 | 978.8 – 1,022.8 | 847 |
| 158 | Claude 3 Haiku (20240307) | Anthropic | 1,000.4 | 987.9 – 1,012.9 | 13,380 |
| 159 | Molmo 7B-D (0924) | allenai | 994.5 | 980.6 – 1,008.3 | 2,815 |
| 160 | Llama 3.2 11B Vision Instruct | Meta | 993.1 | 981.8 – 1,004.5 | 4,817 |
| 161 | nvila-internal-15b-v1 | nvidia | 985.9 | 965.8 – 1,006.0 | 1,077 |
| 162 | LLaVA-OneVision Qwen2 72B OV | 978.6 | 960.4 – 996.8 | 1,321 | |
| 163 | LLaVA 1.6 34B | 965.0 | 952.8 – 977.3 | 4,531 | |
| 164 | CogVLM2 Llama3 Chat 19B | zai | 964.2 | 949.0 – 979.5 | 1,991 |
| 165 | MiniCPM-V 2.6 | 963.0 | 947.2 – 978.9 | 1,987 | |
| 166 | InternVL2 4B | 958.7 | 946.0 – 971.3 | 3,703 | |
| 167 | Phi-3.5 Vision Instruct | 微软 | 919.7 | 903.9 – 935.5 | 2,592 |
| 168 | Phi-3 Vision 128K Instruct | 微软 | 881.3 | 862.6 – 899.9 | 1,401 |
数据来源:Arena 官方视觉理解。 使用官方公开数据集,按 CC BY 4.0 标注来源。 各榜单独同步;来源未发布新数据时,内容日期保持不变。
竞技场评分与任务、评测方式有关,不同子榜的分数不宜直接比较。 最近检查:2026-11-22 18:05(UTC+8)。