大模型能力排名

更新时间：2026-06-10

模型	总分	专业能力	高难度提示词	编程	数学	创意写作	指令遵循	长文本
claude-opus-4-6-thinking	1	1	1	1	2	1	1	1
claude-opus-4-7-thinking	2	6	3	2	6	2	2	3
claude-opus-4-6	3	2	2	5	4	6	3	2
claude-opus-4-7	4	3	4	3	8	4	5	4
muse-spark	5	33	8	9	27	10	19	25
gemini-3.1-pro-preview	6	8	7	10	5	5	7	7
gemini-3-pro	7	16	9	17	19	3	13	12
claude-opus-4-8-thinking	8	18	5	4	16	7	4	5
gpt-5.5-high	9	4	12	14	9	19	8	14
gpt-5.4-high	10	5	11	15	3	26	10	16
claude-opus-4-8	11	7	6	6	11	17	11	6
gemini-3.5-flash	12	9	22	34	1	9	23	21
gpt-5.2-chat-latest-20260210	13	27	16	18	38	41	28	29
glm-5.1	14	12	13	8	18	13	16	13
grok-4.20-beta1	15	49	29	32	40	12	36	39
gpt-5.5	16	15	19	40	10	21	15	23
qwen3.7-max-preview	17	11	17	11	7	31	18	10
gemini-3-flash	18	24	20	31	20	14	26	24
grok-4.20-beta-0309-reasoning	19	43	23	25	24	25	37	38
claude-opus-4-5-20251101-thinking-32k	20	13	14	7	23	8	6	9
grok-4.20-multi-agent-beta-0309	21	32	30	19	41	27	40	41
gpt-5.5-instant	22	45	24	26	30	15	30	30
claude-sonnet-4-6	23	14	10	12	31	20	9	8
ernie-5.1	24	40	25	20	17	35	24	34
claude-opus-4-5-20251101	25	17	15	13	29	11	12	11
grok-4.1-thinking	26	53	37	44	47	45	56	54
gpt-5.4	27	22	26	27	34	38	22	22
qwen3.5-max-preview	28	21	21	24	26	24	17	20
mimo-v2.5-pro	29	10	18	21	14	34	14	15
kimi-k2.6	30	20	27	29	13	39	27	26
gemini-3-flash (thinking-minimal)	31	55	39	51	35	22	42	40
qwen3.6-max-preview	32	23	32	33	15	37	35	31
grok-4.1	33	68	43	53	66	40	57	52
deepseek-v4-pro-thinking	34	46	35	48	21	30	33	27
glm-5	35	30	36	47	56	18	39	33
deepseek-v4-pro	36	44	40	39	59	33	32	32
dola-seed-2.0-pro	37	39	33	28	42	67	51	51
claude-sonnet-4-5-20250929-thinking-32k	38	19	28	16	37	23	20	17
claude-sonnet-4-5-20250929	39	29	31	22	74	16	21	19
gpt-5.1-high	40	38	42	54	36	43	38	42
gemma-4-31b	41	37	44	45	25	50	31	36
kimi-k2.5-thinking	42	35	46	37	22	49	44	43
minimax-m3	43	28	49	30	12	62	34	45
ernie-5.0-preview-1203	44	72	52	77	104	51	68	78
gpt-5.4-mini-high	45	31	45	42	51	71	49	53
mimo-v2-pro	46	25	41	38	39	46	41	35
claude-opus-4-1-20250805-thinking-16k	47	36	34	23	48	28	25	18
gpt-5.3-chat-latest	48	52	47	46	77	65	53	47
ernie-5.0-0110	49	81	51	52	60	47	64	73
claude-opus-4-1-20250805	50	58	38	35	65	32	29	28
grok-4.3	51	82	61	56	80	42	71	59
gemini-2.5-pro	52	60	62	86	50	29	45	44
gpt-4.5-preview-2025-02-27	53	102	86	95	102	36	48	66
qwen3.6-plus	54	47	48	50	33	64	47	48
qwen3.5-397b-a17b	55	34	50	49	44	55	52	46
chatgpt-4o-latest-20250326	56	101	65	80	110	48	60	72
glm-4.7	57	79	53	60	70	63	59	49
gpt-5.1	58	66	67	74	79	57	61	65
gemma-4-26b-a4b	59	48	55	64	28	69	46	55
gpt-5.2-high	60	42	58	55	32	92	62	79
deepseek-v4-flash-thinking	61	54	64	70	54	61	54	56
gpt-5.2	62	56	57	61	63	86	65	69
qwen3-max-preview	63	51	59	63	57	80	63	64
longcat-flash-chat-2602-exp	64	57	60	41	62	89	80	77
deepseek-v4-flash	65	62	63	65	58	60	58	57
gpt-5-high	66	65	78	82	64	110	89	112
gemini-3.1-flash-lite-preview	67	77	81	100	55	54	87	86
mimo-v2.5	68	41	54	57	52	74	55	50
kimi-k2.5-instant	69	67	56	36	49	87	50	61
o3-2025-04-16	70	85	88	97	45	99	100	116
grok-4-1-fast-reasoning	71	87	84	91	86	53	104	95
kimi-k2-thinking-turbo	72	61	69	59	53	83	73	82
amazon-nova-experimental-chat-26-02-10	73	26	70	58	71	145	72	88
mimo-v2-omni	74	63	68	62	43	98	81	60
gpt-5-chat	75	75	71	89	94	84	74	75
glm-4.6	76	86	83	94	82	68	79	80
mistral-medium-3.5	77	90	82	78	46	91	78	84
deepseek-v3.2-exp-thinking	78	76	77	72	72	78	76	85
deepseek-v3.2	79	74	76	79	69	72	66	71
claude-opus-4-20250514-thinking-16k	80	80	66	43	83	44	43	37
qwen3-max-2025-09-23	81	113	73	71	68	79	82	81
qwen3-235b-a22b-instruct-2507	82	71	72	75	84	105	77	76
deepseek-v3.2-exp	83	105	74	84	87	56	75	68
deepseek-v3.2-thinking	84	69	80	69	76	85	69	67
deepseek-r1-0528	85	111	94	87	124	81	114	118
grok-4-fast-chat	86	107	95	96	78	94	106	98
ernie-5.0-preview-1022	87	92	105	132	118	59	105	100
nvidia-nemotron-3-ultra-550b-a55b-nvfp4	88	64	98	108	–	95	108	90
kimi-k2-0905-preview	89	116	92	81	89	101	115	123
deepseek-v3.1	90	103	97	112	91	90	97	99
deepseek-v3.1-terminus-thinking	91	–	79	90	100	96	67	63
kimi-k2-0711-preview	92	120	101	93	134	115	135	132
deepseek-v3.1-thinking	93	99	91	102	92	66	70	62
qwen3.5-122b-a10b	94	83	99	99	81	118	92	106
amazon-nova-experimental-chat-26-01-10	95	50	75	66	101	131	98	96
deepseek-v3.1-terminus	96	–	109	123	126	58	112	102
qwen3-vl-235b-a22b-instruct	97	73	87	85	99	123	84	93
mistral-large-3	98	112	100	83	115	109	96	107
minimax-m2.7	99	84	89	68	96	126	90	83
hunyuan-hy3-preview	100	78	85	92	75	122	95	89