🔬 과학 추론 랭킹

대학원급 과학 추론: GPQA, HLE, SciCode.

🏆 종합673 💻 코딩392 📐 수학32 🧠 추론391 🤖 에이전트31 📚 지식388 👁️ 멀티모달370 🛠️ 코딩 에이전트40 🔬 과학 추론45 🌍 다국어12 ✍️ 글쓰기53 💬 채팅 어시스턴트53 🔭 리서치48 🔧 도구 호출31 📜 긴 컨텍스트55 ⚡ 속도378 💰 가성비377 🔓 오픈 웨이트153

45개 모델

#	모델명	Podium Score	GPQA Diamond	Humanity’s Last Exam	SciCode	속도
01	Seed 2.0 Pro ByteDance	88.9	88.9%	—	—	40 t/s
02	GPT-5.1 OpenAI	88.1	88.1%	—	—	60 t/s
03	Claude Opus 4.5 Anthropic	87.0	87%	—	—	50 t/s
4	Qwen3.5 122B-A10B Alibaba · 오픈소스	86.6	86.6%	—	—	126 t/s
5	GPT-5 OpenAI	85.7	85.7%	—	—	60 t/s
6	Claude Mythos Preview Anthropic · preview	79.7	94.6%	64.7%	—	80 t/s
7	Claude Opus 4.7 Anthropic	74.5	94.2%	54.7%	—	80 t/s
8	Grok 4 Heavy xAI · preview	69.6	88.4%	50.7%	—	50 t/s
9	Claude Sonnet 4.6 Anthropic	69.5	89.9%	49%	—	50 t/s
10	Kimi K3 Moonshot AI · 오픈소스	69.4	93.5%	56%	58.7%	37 t/s
11	Claude Fable 5 Anthropic	68.7	92.6%	53.3%	60.2%	71 t/s
12	Claude Opus 4.8 Anthropic	68.3	93.6%	57.9%	53.5%	50 t/s
13	Gemini 3.1 Pro Google	68.2	94.3%	51.4%	59%	136 t/s
14	Qwen3.8 Max Alibaba	68.1	92.6%	43.6%	—	55 t/s
15	GPT-5.5 OpenAI	67.3	93.6%	52.2%	56.1%	60 t/s
16	Claude Opus 5 Anthropic	67.2	93.2%	52.6%	55.7%	60 t/s
17	Gemini 3 Flash Google	67.0	90.4%	43.5%	—	120 t/s
18	Muse Spark Meta	66.5	89.5%	58.4%	51.5%	80 t/s
19	GPT-5.4 OpenAI	66.3	92.8%	39.8%	—	60 t/s
20	GPT-5.6 Sol OpenAI	66.0	94.6%	47.2%	56.1%	72 t/s
21	GLM-5.2 Z.ai · 오픈소스	65.5	91.2%	54.7%	50.5%	164 t/s
22	Gemini 3 Pro Google	64.6	91.9%	45.8%	56.1%	120 t/s
23	Muse Spark 1.1 Meta	64.4	89.8%	45.1%	58.2%	208 t/s
24	GPT-5.2 OpenAI	63.5	92.4%	34.5%	—	90 t/s
25	Claude Opus 4.6 Anthropic	63.4	91.3%	53.1%	45.7%	50 t/s
26	GPT-5.6 Terra OpenAI	62.9	92.9%	41.8%	53.9%	127 t/s
27	Grok 4.5 xAI	62.5	93%	40.3%	54.1%	59 t/s
28	Qwen3.7 Max Alibaba	62.4	92.4%	41.4%	53.5%	203 t/s
29	Gemini 3.5 Flash Google	62.1	92.2%	41%	53.1%	267 t/s
30	GPT-5.3 Codex OpenAI	61.5	91.5%	39.9%	53.2%	118 t/s
31	Gemini 3.6 Flash Google	61.3	92.8%	38.3%	52.7%	233 t/s
32	DeepSeek V4 Flash DeepSeek · 오픈소스	61.0	88.1%	45.1%	49.9%	113 t/s
33	GPT-5.6 Luna OpenAI	60.7	92.3%	37.2%	52.5%	175 t/s
34	DeepSeek V4 Pro DeepSeek · 오픈소스	60.2	90.1%	48.2%	42.4%	66 t/s
35	Kimi K2.6 Moonshot AI · 오픈소스	59.7	90.5%	36.4%	52.2%	40 t/s
36	Qwen3.7 Plus Alibaba	58.8	90.3%	34.7%	51.3%	54 t/s
37	Qwen3.5 397B-A17B Alibaba · 오픈소스	58.6	88.4%	28.7%	—	71 t/s
38	MiniMax M3 MiniMax	58.5	92.9%	37.1%	45.4%	93 t/s
39	MiMo V2.5 Pro Xiaomi · 오픈소스	56.9	86.6%	33.8%	50.2%	65 t/s
40	Kimi K2.7 Code Moonshot AI · 오픈소스	56.6	89.6%	32.8%	47.5%	39 t/s
41	Hunyuan Hy3 Tencent	56.5	90.4%	31.6%	47.6%	69 t/s
42	Inkling Thinking Machines	54.3	87.2%	29.7%	46.1%	81 t/s
43	Qwen3.6 Plus Alibaba	53.3	90.4%	28.8%	40.7%	55 t/s
44	Nemotron 3 Ultra 550B NVIDIA · 오픈소스	51.2	87%	26.6%	39.9%	142 t/s
45	Claude Sonnet 5 Anthropic	48.8	80%	17.8%	48.6%	96 t/s

Podium Score는 5개 리더보드의 가중 평균입니다. 자세한 내용은 평가 방법을 참고하세요.