排行榜/Arena 人类盲测/Vision 视觉

Arena Vision 视觉排行榜

图像理解 / 多模态问答 Elo 排名 · 完整榜单 165 个模型

Rank	区间	Model	厂商	License	Elo 分	投票数	输入 $/M	输出 $/M	上下文
1	1–10	claude-fable-5-high	Anthropic	Proprietary	1308±7.1	14.2k	$10	$50	1M
2	1–19	qwen3.8-max	Alibaba	Proprietary	1300.7±7.4	10.0k	$2	$6	1M
3	1–19	claude-opus-4-6-high	Anthropic	Proprietary	1299.2±6.5	22.3k	$5	$25	1M
4	1–22	claude-opus-4-7	Anthropic	Proprietary	1298.2±6.6	23.2k	$5	$25	1M
5	1–22	claude-opus-4-7-high	Anthropic	Proprietary	1297.8±6.5	22.7k	$5	$25	1M
6	1–26	gemini-3.7-flash-high	Google	Proprietary	1296.9±9.9	4.5k	$0.75	$3.8	1M
7	2–25	claude-opus-4-6	Anthropic	Proprietary	1294.5±6.4	26.9k	$5	$25	1M
8	1–37	gpt-6.1-sol-max	OpenAI	Proprietary	1294.1±14.5	1.8k	$2	$10	1M
9	1–31	muse-spark	Meta	Proprietary	1293.9±9.1	5.9k	-	-	-
10	1–37	muse-spark-1.2 (xHigh)	Meta	Proprietary	1292.9±14.3	2.0k	$1.3	$4.3	-
11	1–36	claude-opus-5.5-high	Anthropic	Proprietary	1292.7±12.7	2.5k	$4	$20	1M
12	2–36	muse-spark-1.3-max	Meta	Proprietary	1289.7±9.8	4.5k	-	-	-
13	2–34	gemini-3-pro	Google	Proprietary	1289.3±7.4	13.6k	$2	$12	1M
14	2–37	gemini-3.8-flash-high	Google	Proprietary	1289.1±10.6	3.8k	$0.75	$3.8	1M
15	2–35	claude-opus-5-high	Anthropic	Proprietary	1287.4±6.8	16.7k	$5	$25	1M
16	2–37	gpt-5.6-sol-xhigh	OpenAI	Proprietary	1287±7.5	10.9k	$4	$20	-
17	2–38	claude-fable-5.1-max	Anthropic	Proprietary	1286.4±9.7	4.6k	$10	$50	1M
18	2–38	muse-spark-1.3 (xHigh)	Meta	Proprietary	1286.2±10.3	4.1k	-	-	-
19	2–39	gpt-6-astra-max	OpenAI	Proprietary	1286±10.8	3.6k	-	-	-
20	4–36	gpt-5.5	OpenAI	Proprietary	1285.8±5.8	28.0k	$5	$30	1M
21	4–37	claude-opus-4-8-high	Anthropic	Proprietary	1285.6±6.5	18.3k	$5	$25	1M
22	4–37	gemini-3.5-flash-medium	Google	Proprietary	1285.3±6.9	13.2k	$0.75	$4.5	1M
23	6–37	gpt-5.4-high	OpenAI	Proprietary	1285±5.8	29.9k	$2.5	$15	1M
24	6–38	gemini-3.5-flash-high	Google	Proprietary	1283.7±7	13.1k	$0.75	$4.5	1M
25	8–39	gpt-5.5-high	OpenAI	Proprietary	1281.1±5.9	26.3k	$5	$30	1M
26	6–40	gemini-3.6-flash-high	Google	Proprietary	1280.7±8.2	8.1k	$0.75	$3.8	1M
27	7–40	grok-4.5	SpaceXAI	Proprietary	1280±7.2	11.8k	$2	$6	500K
28	8–40	muse-spark-1.1	Meta	Proprietary	1280±7	12.0k	$1.3	$4.3	1M
29	8–39	gemini-3.1-pro-preview	Google	Proprietary	1279.8±5.2	46.3k	$1	$6	1M
30	8–40	claude-opus-4-8	Anthropic	Proprietary	1279±6.5	18.9k	$5	$25	1M
31	8–40	gpt-5.4	OpenAI	Proprietary	1278.9±6.5	22.8k	$1.3	$7.5	1M
32	9–41	gpt-5.2-chat-latest-20260210	OpenAI	Proprietary	1277.9±6.5	16.5k	$1.8	$14	128K
33	9–47	gpt-5.5-instant	OpenAI	Proprietary	1275.7±8.8	7.6k	$5	$30	1M
34	10–45	claude-sonnet-4-6	Anthropic	Proprietary	1275±6.1	27.4k	$1.5	$7.5	1M
35	11–47	glm-5.3-flash	Z.ai	MIT	1273.3±7.1	11.2k	$0.06	$0.20	1M
36	21–47	gemini-3-flash	Google	Proprietary	1272.8±5.2	39.1k	$0.50	$3	1M
37	9–55	gpt-6-sol-max	OpenAI	Proprietary	1270.4±12.3	2.7k	-	-	-
38	24–52	gpt-5.6-terra-xhigh	OpenAI	Proprietary	1268±7.5	11.0k	$2	$12	-
39	14–60	claude-sonnet-5.5-xhigh	Anthropic	Proprietary	1266.2±13.7	2.1k	$2	$10	1M
40	27–55	gemini-3.5-flash-lite	Google	Proprietary	1266±8.1	8.4k	$0.15	$1.3	1M
41	32–55	kimi-k2.6	Moonshot	Modified MIT	1265.4±6.9	16.4k	$0.95	$4	262K
42	33–56	qwen3.7-plus	Alibaba	Proprietary	1264.2±6.8	13.6k	$0.32	$1.3	1M
43	33–58	claude-sonnet-5-high	Anthropic	Proprietary	1263.3±7	13.5k	$2	$10	1M
44	34–58	gemma-4-31b	Google	Apache 2.0	1262.1±5.8	39.7k	$0.99	$1.5	131K
45	33–58	grok-4.6-high	SpaceXAI	Proprietary	1261.9±8.6	6.5k	$2	$6	500K
46	33–62	deepseek-v4.1-flash-max	DeepSeek	MIT	1260.3±10.9	3.6k	$0.30	$1.2	1M
47	37–58	gemini-3-flash (thinking-minimal)	Google	Proprietary	1259.9±5.4	37.2k	$0.50	$3	1M
48	37–62	gpt-5.6-luna-xhigh	OpenAI	Proprietary	1257.7±7.4	11.1k	$0.20	$1.2	-
49	37–63	dola-seed-2.0-pro	Bytedance	Proprietary	1256.1±7.3	11.0k	-	-	-
50	37–62	grok-4.20-beta-0309-reasoning	SpaceXAI	Proprietary	1255.6±6.1	26.8k	$1.3	$2.5	1M
51	41–67	kimi-k2.5-thinking	Moonshot	Modified MIT	1252.1±5.7	28.3k	$0.60	$3	-
52	38–68	gpt-5.1-high	OpenAI	Proprietary	1251.4±7.9	9.9k	$0.63	$5	400K
53	42–67	gpt-5.4-mini-high	OpenAI	Proprietary	1250.9±6.3	25.7k	$0.75	$4.5	400K
54	42–68	grok-4.20-multi-agent-beta-0309	SpaceXAI	Proprietary	1250.5±6.1	24.5k	$1.3	$2.5	1M
55	34–77	Step 5 Preview	StepFun	Proprietary	1250.3±18.2	1.1k	$1	$2.7	1M
56	38–74	mimo-v2.6-flash	Xiaomi	MIT	1247.6±11.9	2.9k	$0.14	$0.28	1M
57	47–68	gemini-2.5-pro	Google	Proprietary	1247.6±4.6	88.8k	$1.3	$10	1M
58	38–74	grok-4.7-xhigh	SpaceXAI	Proprietary	1247.5±11.8	2.9k	$2	$6	500K
59	37–77	mimo-v2.6-pro	Xiaomi	MIT	1246±14.6	1.8k	$0.43	$0.87	1M
60	47–69	qwen3.5-397b-a17b	Alibaba	Apache 2.0	1245.9±5.4	32.1k	$0.50	$3.6	262K
61	46–74	qwen3.8-27b	Alibaba	Apache 2.0	1244.5±8.5	7.0k	$0.40	$3	262K
62	50–72	gpt-5.2-high	OpenAI	Proprietary	1243±6.1	19.0k	$0.88	$7	400K
63	51–72	grok-4.3	SpaceXAI	Proprietary	1242.5±5.9	27.0k	$1.3	$2.5	1M
64	51–74	gemma-4-26b-a4b	Google	Apache 2.0	1242±6.5	23.9k	-	-	-
65	46–78	gpt-6-luna-max	OpenAI	Proprietary	1241±12.1	2.8k	-	-	-
66	51–77	gpt-5.1	OpenAI	Proprietary	1239.6±7.5	10.8k	$0.63	$5	400K
67	53–75	chatgpt-4o-latest-20250326	OpenAI	Proprietary	1238.9±5.5	23.4k	$5	$15	128K
68	51–78	kimi-k2.5-instant	Moonshot	Modified MIT	1236.4±10.8	4.1k	$0.50	$2.5	262K
69	56–78	minimax-m3	MiniMax	MiniMax Community License	1235.5±6.5	19.0k	$0.30	$1.2	524K
70	57–78	gemini-3.1-flash-lite-preview	Google	Proprietary	1234.6±5.8	27.7k	$0.25	$1.5	1M
71	57–78	mimo-v2.5	Xiaomi	MIT	1233.6±5.9	28.8k	$0.14	$0.28	1M
72	59–78	glm-5v-turbo	Z.ai	Proprietary	1230.7±5.6	42.2k	$1.2	$4	203K
73	57–86	gemini-2.5-flash-preview-09-2025	Google	Proprietary	1229.3±9.8	4.8k	$0.30	$2.5	1M
74	63–81	gpt-5.2	OpenAI	Proprietary	1228.7±6	20.1k	$0.88	$7	400K
75	64–86	qwen3.5-122b-a10b	Alibaba	Apache 2.0	1226.3±6.8	13.2k	$0.26	$2.1	262K
76	64–86	gpt-5-chat	OpenAI	Proprietary	1225.8±7	39.9k	$5	$30	-
77	59–88	gpt-4.5-preview-2025-02-27	OpenAI	Proprietary	1224.9±11.7	2.9k	$75	$150	128K
78	67–89	ernie-5.0-preview-1220	Baidu	Proprietary	1219.4±11.1	3.4k	-	-	-
79	74–88	qwen3.5-27b	Alibaba	Apache 2.0	1216.9±5.6	29.6k	$0.26	$2.6	262K
80	73–89	mimo-v2-omni	Xiaomi	Proprietary	1216.1±7.4	10.7k	$0.44	$2.2	256K
81	74–89	o3-2025-04-16	OpenAI	Proprietary	1215.6±6.6	45.6k	$2	$8	200K
82	74–88	gemini-2.5-flash	Google	Proprietary	1215.2±4.6	60.1k	$0.30	$2.5	1M
83	74–91	qwen3-vl-235b-a22b-instruct	Alibaba	Apache 2.0	1215±7	12.8k	$0.29	$1.1	262K
84	74–93	gpt-4.1-2025-04-14	OpenAI	Proprietary	1214.2±6.7	41.0k	$2	$8	1M
85	77–96	gpt-5-high	OpenAI	Proprietary	1210±7	34.5k	$0.63	$5	400K
86	77–99	Inkling Small	Thinky	Apache 2.0	1207.6±8.4	7.5k	$0.45	$1.2	524K
87	73–102	claude-opus-4-20250514-thinking-16k	Anthropic	Proprietary	1207.5±16	1.3k	$15	$75	200K
88	73–103	claude-sonnet-4-20250514-thinking-32k	Anthropic	Proprietary	1206.2±16.9	1.1k	$3	$15	200K
89	83–101	gpt-4.1-mini-2025-04-14	OpenAI	Proprietary	1200.7±7.6	40.2k	$0.40	$1.6	1M
90	84–101	o4-mini-2025-04-16	OpenAI	Proprietary	1200.7±7.1	41.3k	$1.1	$4.4	200K
91	80–103	mistral-medium-3.5	Mistral	Modified MIT	1200.4±9.6	5.4k	$0.75	$3.8	262K
92	84–101	mistral-large-3	Mistral	Apache 2.0	1199.8±8	8.6k	$0.50	$1.5	-
93	85–101	gpt-5.4-nano-high	OpenAI	Proprietary	1199.4±6.4	25.6k	$0.20	$1.3	400K
94	85–103	grok-4-1-fast-reasoning	SpaceXAI	Proprietary	1196.3±6.9	13.4k	$0.20	$0.50	128K
95	83–107	claude-3-7-sonnet-20250219-thinking-32k	Anthropic	Proprietary	1192.6±15.5	1.4k	$3	$15	200K
96	85–106	o1-2024-12-17	OpenAI	Proprietary	1192.6±10.8	3.7k	$15	$60	200K
97	86–107	qwen3-vl-235b-a22b-thinking	Alibaba	Apache 2.0	1188.6±12.4	2.4k	$0.40	$4	131K
98	87–106	gemini-2.5-flash-lite-preview-06-17-thinking	Google	Proprietary	1188.5±7.5	35.9k	$0.10	$0.40	1M
99	86–107	claude-opus-4-20250514	Anthropic	Proprietary	1188.5±12.9	2.2k	$5	$25	200K
100	86–108	claude-sonnet-4-20250514	Anthropic	Proprietary	1188.3±14	1.8k	$3	$15	200K
101	87–108	qwen-vl-max-2025-08-13	Alibaba	Proprietary	1185.3±12.1	3.2k	$0.52	$2.1	131K
102	92–107	grok-4-0709	SpaceXAI	Proprietary	1184±7.4	30.2k	$1.3	$2.5	256K
103	91–107	gpt-5-mini-high	OpenAI	Proprietary	1183.4±8.1	29.0k	$0.13	$1	400K
104	95–110	gemini-1.5-pro-002	Google	Proprietary	1179.3±9.1	8.9k	$3.5	$10.5	2M
105	95–114	claude-3-7-sonnet-20250219	Anthropic	Proprietary	1175±9.4	4.5k	$3	$15	200K
106	95–115	gemini-2.5-flash-lite-preview-09-2025-no-thinking	Google	Proprietary	1174.6±9.6	4.9k	$0.10	$0.40	1M
107	97–116	gemini-2.0-flash-001	Google	Proprietary	1171.8±7.2	10.1k	$0.10	$0.40	1M
108	104–121	gpt-4o-2024-05-13	OpenAI	Proprietary	1161.6±8.8	23.3k	$5	$15	128K
109	102–123	glm-4.6v	Z.ai	MIT	1161±13.8	2.5k	$0.30	$0.90	131K
110	105–121	claude-3-5-sonnet-20241022	Anthropic	Proprietary	1160.4±7.9	10.4k	$3	$15	200K
111	105–122	gemma-3-27b-it	Google	Gemma	1159.5±8.1	17.0k	$0.08	$0.45	131K
112	106–121	mistral-medium-2508	Mistral	Proprietary	1159.3±6.3	41.6k	$0.40	$2	131K
113	104–125	step-1o-turbo-202506	StepFun	Proprietary	1156±14.3	1.9k	-	-	-
114	108–123	mistral-medium-2505	Mistral	Proprietary	1155.2±8.3	10.8k	$0.40	$2	131K
115	105–124	hunyuan-vision-1.5-thinking	Tencent	Proprietary	1155±12.3	2.4k	-	-	-
116	107–125	glm-4.5v	Z.ai	MIT	1152.9±11.8	3.2k	$0.60	$1.8	66K
117	105–129	hunyuan-large-vision	Tencent	Proprietary	1149.3±16.5	1.3k	-	-	-
118	108–128	gpt-5-nano-high	OpenAI	Proprietary	1146.3±11.4	3.9k	$0.03	$0.20	400K
119	108–127	llama-4-maverick-17b-128e-instruct	Meta	Llama 4	1145.3±9.2	6.9k	$0.63	$1.8	131K
120	108–129	step-3	StepFun	Apache 2.0	1145±11.9	3.2k	$0.57	$1.4	66K
121	108–128	claude-3-5-sonnet-20240620	Anthropic	Proprietary	1144.8±10.1	21.6k	$3	$15	200K
122	111–131	gemini-1.5-flash-002	Google	Proprietary	1141.5±9.9	7.2k	$0.07	$0.30	1M
123	112–132	mistral-small-2506	Mistral	Apache 2.0	1139.8±9.1	10.8k	$0.10	$0.30	262K
124	114–133	gemini-2.0-flash-lite-preview-02-05	Google	Proprietary	1135.8±10.4	4.0k	$0.07	$0.30	1M
125	117–134	llama-4-scout-17b-16e-instruct	Meta	Llama	1127.4±9.5	6.5k	$0.05	$0.10	16K
126	118–134	mistral-small-3.1-24b-instruct-2503	Mistral	Apache 2.0	1127.2±8.5	28.5k	$0.10	$0.30	32K
127	115–134	claude-3-5-haiku-20241022	Anthropic	Proprietary	1126.6±16.1	1.4k	$1	$5	200K
128	117–134	step-1o-vision-32k-highres	StepFun	Proprietary	1124.2±12.7	2.8k	-	-	-
129	122–134	qwen2.5-vl-72b-instruct	Alibaba	Qwen	1121±10.9	3.8k	$0.80	$1	128K
130	123–134	gemini-1.5-pro-001	Google	Proprietary	1119.5±11.6	16.7k	$3.5	$10.5	2M
131	120–137	qwen2.5-vl-32b-instruct	Alibaba	Apache 2.0	1118.9±15.5	1.5k	$0.20	$0.60	128K
132	124–135	gpt-4o-2024-08-06	OpenAI	Proprietary	1118±12.3	3.4k	$2.5	$10	128K
133	122–139	molmo-2-8b	Ai2	Apache 2.0	1112.1±19.8	1.2k	$0.20	$0.20	37K
134	125–138	gpt-4-turbo-2024-04-09	OpenAI	Proprietary	1111.8±11.5	13.4k	$10	$30	128K
135	132–139	gpt-4o-mini-2024-07-18	OpenAI	Proprietary	1096.9±8.3	17.3k	$0.15	$0.60	128K
136	132–139	pixtral-large-2411	Mistral	MRL	1095.2±9.3	5.4k	$2	$6	128K
137	131–143	gpt-4.1-nano-2025-04-14	OpenAI	Proprietary	1088.5±18.3	1.2k	$0.10	$0.40	1M
138	133–143	qwen-vl-max-1119	Alibaba	Proprietary	1084.5±16	1.4k	$0.80	$3.2	131K
139	134–141	qwen2-vl-72b	Alibaba	Qwen	1084.5±10.2	5.9k	$0.90	$0.90	33K
140	137–144	gemini-1.5-flash-8b-001	Google	Proprietary	1071.2±10.6	6.2k	$0.07	$0.30	1M
141	137–145	step-1v-32k	StepFun	Proprietary	1064±16.8	1.5k	-	-	-
142	138–145	claude-3-opus-20240229	Anthropic	Proprietary	1061.3±10.4	15.6k	$15	$75	200K
143	138–145	gemini-1.5-flash-001	Google	Proprietary	1059.2±11.9	13.3k	$0.07	$0.30	1M
144	141–151	molmo-72b-0924	Ai2	Apache 2.0	1044.6±13.1	3.0k	-	-	-
145	140–153	hunyuan-standard-vision-2024-12-31	Tencent	Proprietary	1041.4±21	809	-	-	-
146	144–152	llama-3.2-vision-90b-instruct	Meta	Llama 3.2	1032.1±9.3	8.7k	-	-	-
147	144–154	qwen2-vl-7b-instruct	Alibaba	Apache 2.0	1030.5±10.5	5.8k	$0.20	$0.20	33K
148	144–154	pixtral-12b-2409	Mistral	Apache 2.0	1025.1±9.5	7.5k	$0.15	$0.15	128K
149	144–155	internvl2-26b	-	MIT	1024.7±12.9	5.1k	-	-	-
150	144–158	amazon-nova-lite-v1.0	Amazon	Proprietary	1018.5±15	1.9k	$0.06	$0.24	300K
151	144–157	amazon-nova-pro-v1.0	Amazon	Proprietary	1018.4±13.8	2.3k	$0.80	$3.2	300K
152	145–157	claude-3-sonnet-20240229	Anthropic	Proprietary	1016.1±11.4	12.3k	$3	$15	200K
153	147–159	yi-vision	-	Proprietary	1002.1±18.1	1.2k	-	-	-
154	146–161	c4ai-aya-vision-32b	Cohere	CC-BY-NC-4.0	1000.1±22	847	-	-	-
155	149–159	claude-3-haiku-20240307	Anthropic	Proprietary	999.6±12.4	13.4k	$0.25	$1.3	200K
156	150–159	molmo-7b-d-0924	Ai2	Apache 2.0	993.5±13.8	2.8k	-	-	-
157	152–159	llama-3.2-vision-11b-instruct	Meta	Llama 3.2	992.6±11.3	4.8k	-	-	-
158	150–163	nvila-internal-15b-v1	Nvidia	-	985±20.1	1.1k	-	-	-
159	153–163	llava-onevision-qwen2-72b-ov	-	Apache 2.0	977.9±18.2	1.3k	-	-	-
160	158–163	llava-v1.6-34b	-	Apache 2.0	964.1±12.3	4.5k	-	-	-
161	157–163	cogvlm2-llama3-chat-19b	Z.ai	CogVLM2	963.5±15.2	2.0k	-	-	-
162	157–163	minicpm-v-2_6	-	Apache 2.0	962.3±15.8	2.0k	-	-	-
163	158–163	internvl2-4b	-	MIT	958.2±12.6	3.7k	-	-	-
164	164–164	phi-3.5-vision-instruct	Microsoft	MIT	919±15.8	2.6k	$0.13	$0.52	128K
165	165–165	phi-3-vision-128k-instruct	Microsoft	MIT	880.6±18.7	1.4k	$0.20	$0.20	32K

Elo 分真人在不知道模型名字的情况下两两投票，胜负累积出的等级分。绝对值没有意义，差值才有：分差 100 分 ≈ 强的那个有 64% 胜率，只差几分基本是平手。
± 后面的数是 95% 置信区间的半宽，投票越多越小。两个模型的区间重叠，说明它们的名次差别在统计上不成立。
区间把上面这个置信区间换算成名次：1–5 意思是这个模型的真实名次可能在第 1 到第 5 之间。

数据更新: 2026-10-14数据来源: Arena AI ↗