最佳写作AI模型

使用Arena Elo评分(归一化到0-100刻度)对AI模型的写作质量进行排名。

Last updated: 35m ago

第一名模型

Claude Opus 4.6

评分: 100

平均评分

81.4

所有排名模型

已排名模型

107

有基准测试数据

权重：Arena Elo (100%)

Top Best for Writing Models by Weighted Score

Top 15 models by weighted score

LMMarketCap.com

#	模型	提供商	评分	Arena Elo
1	Claude Opus 4.6Anthropic	Anthropic	100	100
2	Gemini 3.1 Pro PreviewGoogle	Google	99	99
3	Claude Opus 4.7Anthropic	Anthropic	98.5	98.5
4	Muse Spark 1.1meta	meta	98.5	98.5
5	GPT-5.4OpenAI	OpenAI	97.5	97.5
6	GPT-5.2OpenAI	OpenAI	96.8	96.8
7	MiMo-V2.6-ProXiaomi	Xiaomi	96.7	96.7
8	Claude Opus 4.8Anthropic	Anthropic	96.5	96.5
9	Gemini 3.5 FlashGoogle	Google	96.2	96.2
10	GPT-5.2 ChatOpenAI	OpenAI	96	96
11	GPT-5.5OpenAI	OpenAI	95.8	95.8
12	GPT-5.1OpenAI	OpenAI	95.8	95.8
13	GLM 5.3 FlashZhipu AI	Zhipu AI	95.7	95.7
14	Gemini 3 Flash PreviewGoogle	Google	95.7	95.7
15	MiMo-V2.5-ProXiaomi	Xiaomi	94.5	94.5
16	Grok 4.5xAI	xAI	94.2	94.2
17	GLM 5.1Zhipu AI	Zhipu AI	94.2	94.2
18	GPT-5OpenAI	OpenAI	94.2	94.2
19	Kimi K2.6Moonshot AI	Moonshot AI	93.5	93.5
20	Claude Sonnet 4.6Anthropic	Anthropic	93.3	93.3
21	Qwen3.6 Max PreviewAlibaba	Alibaba	93.3	93.3
22	GLM 5Zhipu AI	Zhipu AI	92.8	92.8
23	Hy3Tencent	Tencent	92.8	92.8
24	Gemini 3.5 Flash LiteGoogle	Google	92.7	92.7
25	Qwen3.7 PlusAlibaba	Alibaba	92.7	92.7
26	MiMo-V2.6-FlashXiaomi	Xiaomi	92.3	92.3
27	Gemma 4 31BGoogle	Google	92.2	92.2
28	Claude Sonnet 4.5Anthropic	Anthropic	92	92
29	Claude Opus 4.1Anthropic	Anthropic	91.7	91.7
30	Gemini 2.5 ProGoogle	Google	90.7	90.7
31	Qwen3.6 PlusAlibaba	Alibaba	90.5	90.5
32	Inklingthinkingmachines	thinkingmachines	90.3	90.3
33	Qwen3.5 397B A17BAlibaba	Alibaba	90.3	90.3
34	GLM 4.7Zhipu AI	Zhipu AI	90.2	90.2
35	MiniMax M3MiniMax	MiniMax	90	90
36	Qwen3.8 27BAlibaba	Alibaba	89.7	89.7
37	Gemma 4 26B A4B Google	Google	89.7	89.7
38	MiMo-V2.5Xiaomi	Xiaomi	89	89
39	GLM 5V TurboZhipu AI	Zhipu AI	88.8	88.8
40	Gemini 3.1 Flash Lite PreviewGoogle	Google	88.8	88.8
41	Claude Opus 4.5Anthropic	Anthropic	88.3	88.3
42	Mistral Medium 3.5Mistral AI	Mistral AI	87.7	87.7
43	DeepSeek V3.2DeepSeek	DeepSeek	87.5	87.5
44	GLM 4.6Zhipu AI	Zhipu AI	87.3	87.3
45	DeepSeek V3.2 ExpDeepSeek	DeepSeek	87	87
46	DeepSeek V3.1DeepSeek	DeepSeek	86.2	86.2
47	Qwen3.5-122B-A10BAlibaba	Alibaba	86	86
48	o3OpenAI	OpenAI	85.8	85.8
49	MiniMax M2.7MiniMax	MiniMax	85.8	85.8
50	DeepSeek V3.1 TerminusDeepSeek	DeepSeek	85.8	85.8
51	Qwen3 VL 235B A22B InstructAlibaba	Alibaba	85.5	85.5
52	GLM 4.5Zhipu AI	Zhipu AI	85.2	85.2
53	Hy3 previewTencent	Tencent	84.8	84.8
54	Qwen3.5-27BAlibaba	Alibaba	84.8	84.8
55	Inkling Smallthinkingmachines	thinkingmachines	84.2	84.2
56	Qwen3 Next 80B A3B InstructAlibaba	Alibaba	83.2	83.2
57	Qwen3.5-FlashAlibaba	Alibaba	82.7	82.7
58	Gemini 2.5 FlashGoogle	Google	82.5	82.5
59	Qwen3 VL 235B A22B ThinkingAlibaba	Alibaba	82.5	82.5
60	Qwen3.5-35B-A3BAlibaba	Alibaba	82.3	82.3
61	Step 3.5 FlashStepFun	StepFun	82.2	82.2
62	MiniMax M2.5MiniMax	MiniMax	81.8	81.8
63	GPT-5 MiniOpenAI	OpenAI	81.7	81.7
64	Claude Sonnet 4Anthropic	Anthropic	81.2	81.2
65	GPT-4.1 MiniOpenAI	OpenAI	80.5	80.5
66	R1 0528DeepSeek	DeepSeek	80	80
67	o4 MiniOpenAI	OpenAI	80	80
68	GLM 4.6VZhipu AI	Zhipu AI	79.8	79.8
69	GLM 4.5 AirZhipu AI	Zhipu AI	78.8	78.8
70	o3 MiniOpenAI	OpenAI	78.5	78.5
71	R1DeepSeek	DeepSeek	78.2	78.2
72	Qwen3 Next 80B A3B ThinkingAlibaba	Alibaba	78.2	78.2
73	Trinity Large Thinkingarcee-ai	arcee-ai	78	78
74	GLM 4.7 FlashZhipu AI	Zhipu AI	77.5	77.5
75	MiniMax M1MiniMax	MiniMax	77.3	77.3
76	o3 Mini HighOpenAI	OpenAI	77.3	77.3
77	GLM 4.5VZhipu AI	Zhipu AI	75.3	75.3
78	gpt-oss-120bOpenAI	OpenAI	75.3	75.3
79	o1OpenAI	OpenAI	75	75
80	Qwen3 8BAlibaba	Alibaba	74.5	74.5
81	DeepSeek V3 0324DeepSeek	DeepSeek	74.2	74.2
82	MiniMax M2MiniMax	MiniMax	73.8	73.8
83	Mercury 2Inception	Inception	73.8	73.8
84	GPT-5 NanoOpenAI	OpenAI	73	73
85	Nova 2 LiteAmazon	Amazon	72.5	72.5
86	Llama 4 MaverickMeta	Meta	70.8	70.8
87	GPT-4.1 NanoOpenAI	OpenAI	70.3	70.3
88	DeepSeek V3DeepSeek	DeepSeek	69.7	69.7
89	GPT-4o-mini (2024-07-18)OpenAI	OpenAI	69.7	69.7
90	gpt-oss-20bOpenAI	OpenAI	69.7	69.7
91	Mistral Large 2407Mistral AI	Mistral AI	69	69
92	GPT-4.1OpenAI	OpenAI	66.7	66.7
93	Granite 4.2 8BIBM	IBM	64.8	64.8
94	GPT-4oOpenAI	OpenAI	64.3	64.3
95	Mistral LargeMistral AI	Mistral AI	63.3	63.3
96	GPT-4OpenAI	OpenAI	62.7	62.7
97	Command ACohere	Cohere	60.2	60.2
98	Phi 4Microsoft	Microsoft	59.3	59.3
99	GPT-4 TurboOpenAI	OpenAI	59.2	59.2
100	Command R+ (08-2024)Cohere	Cohere	58.3	58.3
101	Llama 3.3 70B InstructMeta	Meta	57.2	57.2
102	Claude Haiku 4.5Anthropic	Anthropic	56.7	56.7
103	GPT-4o-miniOpenAI	OpenAI	53.7	53.7
104	Llama 3.1 8B InstructMeta	Meta	51.8	51.8
105	Llama 3.1 70B InstructMeta	Meta	49.7	49.7
106	Llama 3.2 3B InstructMeta	Meta	44.5	44.5
107	Llama 3.2 1B InstructMeta	Meta	35.2	35.2

评分计算方式

每个模型的评分是其可用基准测试结果的加权平均值。当模型缺少某些基准测试时，权重会在可用的基准测试之间重新归一化。所有评分均为0-100的刻度。数据来源于官方模型卡片、已发表论文和第三方评估平台。

其他专业排行榜

编程最佳数学最佳推理最佳指令遵循最佳数据分析最佳角色扮演最佳多语言最佳

Frequently Asked Questions

根据我们的基准测试分析，Anthropic的Claude Opus 4.6目前在Writing领域排名第一，加权评分为100/100。

模型使用Arena Elo基准测试分数的加权平均进行排名。所有分数均归一化到0-100的刻度。

我们目前对107个拥有相关基准测试数据的模型进行了Writing任务排名。