最佳多语言AI模型

使用跨语言MMLU基准测试分数对AI模型的多语言性能进行排名。找到翻译和非英语任务的最佳LLM。

Last updated: just now

第一名模型

Grok 4.20 Beta

评分: 99.3

平均评分

79.3

所有排名模型

已排名模型

有基准测试数据

权重：MMLU (60%)Arena Elo (40%)

#	模型	提供商	评分	MMLU	Arena Elo
1	Grok 4.20 BetaxAI	xAI	99.3	--	99.3
2	Gemini 3.1 Pro PreviewGoogle	Google	98.7	--	98.7
3	GPT-5.2 ChatOpenAI	OpenAI	96.8	--	96.8
4	Grok 4.1 FastxAI	xAI	95.5	--	95.5
5	GPT-5.4OpenAI	OpenAI	95.4	94	97.5
6	Claude Opus 4.6Anthropic	Anthropic	95.3	92.1	100
7	GPT-5.2OpenAI	OpenAI	94.8	93.5	96.8
8	GPT-5OpenAI	OpenAI	93.5	93	94.2
9	GPT-5.1OpenAI	OpenAI	92.7	--	92.7
10	Grok 4xAI	xAI	92.4	91.5	93.7
11	Claude Sonnet 4.6Anthropic	Anthropic	92.1	91.2	93.3
12	Qwen3.5 397B A17BAlibaba	Alibaba	91.7	--	91.7
13	Claude Opus 4.1Anthropic	Anthropic	91.5	--	91.5
14	Claude Sonnet 4.5Anthropic	Anthropic	91.3	90.8	92
15	Gemini 3 Flash PreviewGoogle	Google	91.1	88	95.7
16	Gemini 2.5 ProGoogle	Google	90.7	90.8	90.7
17	Claude Opus 4.5Anthropic	Anthropic	90.2	91.4	88.3
18	o3OpenAI	OpenAI	89.7	92.3	85.8
19	Gemini 3.1 Flash Lite PreviewGoogle	Google	89.5	--	89.5
20	Claude Opus 4Anthropic	Anthropic	89.3	91	86.7
21	GPT-5 ChatOpenAI	OpenAI	87.8	--	87.8
22	DeepSeek V3.2 ExpDeepSeek	DeepSeek	87.3	--	87.3
23	DeepSeek V3.2DeepSeek	DeepSeek	87.2	--	87.2
24	Grok 4 FastxAI	xAI	87	--	87
25	R1 0528DeepSeek	DeepSeek	86.9	91.5	80
26	Qwen3.5-122B-A10BAlibaba	Alibaba	86.5	--	86.5
27	DeepSeek V3.1DeepSeek	DeepSeek	86.5	--	86.5
28	DeepSeek V3.1 TerminusDeepSeek	DeepSeek	86.2	--	86.2
29	Claude Sonnet 4Anthropic	Anthropic	86.2	89.5	81.2
30	Qwen3 VL 235B A22B InstructAlibaba	Alibaba	86	--	86
31	R1DeepSeek	DeepSeek	85.7	90.8	78.2
32	o1OpenAI	OpenAI	85.1	91.8	75
33	Qwen3.5-27BAlibaba	Alibaba	85	--	85
34	Gemini 2.5 FlashGoogle	Google	84.5	85.8	82.5
35	Claude 3.7 SonnetAnthropic	Anthropic	84.4	90.2	75.7
36	MiniMax M2.5MiniMax	MiniMax	84	--	84
37	Qwen3 Next 80B A3B InstructAlibaba	Alibaba	83.7	--	83.7
38	LongCat Flash ChatMeituan	Meituan	83.5	--	83.5
39	o3 MiniOpenAI	OpenAI	83.5	86.9	78.5
40	Qwen3.5-FlashAlibaba	Alibaba	83.3	--	83.3
41	DeepSeek V3 0324DeepSeek	DeepSeek	83.2	89.2	74.2
42	Qwen3.5-35B-A3BAlibaba	Alibaba	83	--	83
43	Qwen3 VL 235B A22B ThinkingAlibaba	Alibaba	82.7	--	82.7
44	Grok 3xAI	xAI	82.6	88.5	73.7
45	GPT-5 MiniOpenAI	OpenAI	81.8	--	81.8
46	Step 3.5 FlashStepFun	StepFun	81.5	--	81.5
47	Claude 3.7 Sonnet (thinking)Anthropic	Anthropic	81.3	--	81.3
48	Llama 4 MaverickMeta	Meta	81.1	88	70.8
49	DeepSeek V3DeepSeek	DeepSeek	81	88.5	69.7
50	GPT-4.1OpenAI	OpenAI	80.2	89.2	66.7
51	o4 MiniOpenAI	OpenAI	80	--	80
52	GPT-4oOpenAI	OpenAI	79	88.7	64.3
53	Qwen3 Next 80B A3B ThinkingAlibaba	Alibaba	78.2	--	78.2
54	Claude 3.5 SonnetAnthropic	Anthropic	78	88.7	61.8
55	MiniMax M1MiniMax	MiniMax	77.8	--	77.8
56	o3 Mini HighOpenAI	OpenAI	77.3	--	77.3
57	Grok 3 Mini BetaxAI	xAI	76.3	--	76.3
58	Mistral LargeMistral AI	Mistral AI	76.2	84.7	63.3
59	Gemini 2.0 FlashGoogle	Google	76	76.4	75.3
60	gpt-oss-120bOpenAI	OpenAI	75.8	--	75.8
61	Command ACohere	Cohere	75.7	--	75.7
62	GPT-4 TurboOpenAI	OpenAI	75.6	86.5	59.2
63	Gemma 2 27BGoogle	Google	75.2	75.2	--
64	MiniMax M2MiniMax	MiniMax	74.7	--	74.7
65	Qwen3 8BAlibaba	Alibaba	74.7	--	74.7
66	Phi 4Microsoft	Microsoft	74.6	84.8	59.3
67	Llama 3.3 70B InstructMeta	Meta	74.6	86.3	57.2
68	GPT-4o (2024-05-13)OpenAI	OpenAI	74.3	--	74.3
69	Llama 3.3 Nemotron Super 49B V1.5NVIDIA	NVIDIA	73.7	--	73.7
70	Claude Haiku 4.5Anthropic	Anthropic	73.4	84.5	56.7
71	GPT-5 NanoOpenAI	OpenAI	73	--	73
72	Nova 2 LiteAmazon	Amazon	73	--	73
73	QwQ 32BAlibaba	Alibaba	72.7	--	72.7
74	GPT-4o (2024-08-06)OpenAI	OpenAI	72.5	--	72.5
75	Olmo 3.1 32B InstructAllen AI	Allen AI	71.8	--	71.8
76	Llama 3.1 70B InstructMeta	Meta	71.5	86	49.7
77	GPT-4o-miniOpenAI	OpenAI	70.7	82	53.7
78	GPT-4o-mini (2024-07-18)OpenAI	OpenAI	69.7	--	69.7
79	gpt-oss-20bOpenAI	OpenAI	69.7	--	69.7
80	Mistral Large 2407Mistral AI	Mistral AI	69	--	69
81	MercuryInception	Inception	68.2	--	68.2
82	Olmo 3 32B ThinkAllen AI	Allen AI	67.8	--	67.8
83	Qwen2.5 72B InstructAlibaba	Alibaba	67.2	--	67.2
84	Claude 3.5 HaikuAnthropic	Anthropic	67.1	80.9	46.3
85	Llama 3.1 Nemotron 70B InstructNVIDIA	NVIDIA	66.5	--	66.5
86	Olmo 3.1 32B ThinkAllen AI	Allen AI	64.2	--	64.2
87	Llama 3 70B InstructMeta	Meta	62.7	--	62.7
88	Qwen2.5 Coder 32B InstructAlibaba	Alibaba	61.8	--	61.8
89	Claude 3 HaikuAnthropic	Anthropic	60.2	--	60.2
90	Command R (08-2024)Cohere	Cohere	58.3	--	58.3
91	Llama 3 8B InstructMeta	Meta	53.8	--	53.8
92	Llama 3.1 8B InstructMeta	Meta	52	--	52
93	Llama 3.2 3B InstructMeta	Meta	44.5	--	44.5
94	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写作最佳指令遵循最佳数据分析最佳角色扮演最佳

Frequently Asked Questions

根据我们的基准测试分析，xAI的Grok 4.20 Beta目前在Multilingual领域排名第一，加权评分为99.3/100。

模型使用MMLU、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94个拥有相关基准测试数据的模型进行了Multilingual任务排名。