LLM 리더보드

10개 벤치마크에 걸친 25개 모델의 종합 순위.

All Chat Code Reasoning Image Video Agent

25 25개 모델

#	모델	종합	ELO	MMLU	GPQA	HumanEval	SWE-Bench	속도	가격
1	Claude Opus 4 Anthropic	97.8	1385	92.1%	83.5%	94.5%	72.5%	45 t/s	$75/M
2	Gemini 2.5 Pro Google	97.5	1398	92%	84%	93%	63.8%	85 t/s	$10/M
3	o3 OpenAI	97.1	1380	91.2%	81.3%	92.7%	71.7%	58 t/s	$40/M
4	DeepSeek R1 DeepSeek 오픈	92.3	1358	90.8%	71.5%	92.1%	49.2%	32 t/s	$2.19/M
5	o4-mini OpenAI	91.5	1345	88.3%	72.1%	91.5%	68.1%	145 t/s	$4.4/M
6	Claude Sonnet 4 Anthropic	90.3	1340	90.4%	70.2%	93.7%	62.3%	79 t/s	$15/M
7	GPT-4o OpenAI	89.2	1287	88.7%	53.6%	90.2%	38.4%	110 t/s	$10/M
8	Grok 3 xAI	88.8	1330	89.2%	68.5%	90.5%	55.2%	55 t/s	$15/M
9	Claude 3.5 Sonnet Anthropic	88.4	1272	88.7%	65%	92%	49%	82 t/s	$15/M
10	Llama 4 Maverick Meta 오픈	87.5	1325	89.5%	62.8%	90.8%	47.5%	120 t/s	$1.2/M
11	Gemini 2.5 Flash Google	86.7	1310	86.5%	62.3%	89.8%	48.2%	203 t/s	$0.6/M
12	DeepSeek V3 DeepSeek 오픈	86.1	1318	88.5%	59.1%	89.4%	42%	68 t/s	$1.1/M
13	Qwen 3 235B Alibaba 오픈	84.5	1305	87.8%	58.3%	88.7%	41.3%	95 t/s	$1.2/M
14	Llama 3.1 405B Meta 오픈	82.1	1265	87.3%	51.1%	89%	35.2%	29 t/s	$1/M
15	Gemini 2.0 Flash Google	80.5	1290	85.1%	55.4%	87.6%	36.8%	258 t/s	$0.4/M
16	Llama 3.3 70B Meta 오픈	79.8	1258	86%	48.7%	86.4%	32.8%	65 t/s	$0.4/M
17	Claude 3.5 Haiku Anthropic	79.2	1248	84.1%	48.2%	88.1%	40.6%	148 t/s	$4/M
18	Mistral Large Mistral AI	76.5	1245	84%	46.5%	84.3%	30.1%	48 t/s	$6/M
19	Qwen 2.5 72B Alibaba 오픈	76	1230	85.2%	45.6%	82.4%	26.5%	55 t/s	$0.4/M
20	GPT-4o Mini OpenAI	75.8	1271	82%	41.1%	87.2%	23.4%	167 t/s	$0.6/M
21	Grok 2 xAI	74.8	1240	83.2%	44.8%	82.1%	28.3%	60 t/s	$10/M
22	Llama 3.1 70B Meta 오픈	74.3	1235	83.7%	42.3%	80.5%	27.3%	62 t/s	$0.4/M
23	Mixtral 8x22B Mistral AI 오픈	65.2	1195	77.8%	33.5%	75.8%	18.2%	82 t/s	$0.5/M
24	Phi-4 Microsoft	64.5	1150	78.5%	38.5%	72.4%	12.5%	125 t/s	$0.14/M
25	Command R+ Cohere	60.8	1185	75.6%	30.2%	68.2%	15.8%	52 t/s	$10/M