Humanity's Last Exam

数学、科学、人文科学にわたる2,500の専門家検証済み問題からなるフロンティアレベルのベンチマーク。最終的な閉形式の学術評価として設計されている。

モデル: 267 現在、 684 合計
中央値: 12.0%
範囲: 1.1%–74.3%

スコア分布

最低: 1.1%
中央値: 12.0%
最高: 74.3%

リーダーボード

非推奨のモデルを含める

ランク	モデル	スコア
1	Claude Fable 5.2 (Max, Default Fallback) Anthropic	74.3%
2	Claude Fable 5.2 (Xhigh, Default Fallback) Anthropic	68.9%
3	Claude Fable 5.2 (High, Default Fallback) Anthropic	66.2%
4	Claude Fable 5.2 (Medium, Default Fallback) Anthropic	61.9%
5	Claude Opus 5.5 (Max, Default Fallback) Anthropic	61.4%
6	Claude Fable 5.2 (Low, Default Fallback) Anthropic	59.7%
7	Claude Opus 5.5 (Xhigh, Default Fallback) Anthropic	57.5%
8	Gemini 4 Argon (High) Google	57.1%
9	Claude Opus 5.5 (High, Default Fallback) Anthropic	55.6%
10	Claude Sonnet 5.5 (Max, Default Fallback) Anthropic	55.0%
11	Claude Opus 5.5 (Medium, Default Fallback) Anthropic	54.7%
12	GPT-6 Astra (Max) OpenAI	54.7%
13	GPT-6 Astra (Xhigh) OpenAI	54.6%
14	Muse Wonder Full (Max) Meta	53.6%
15	GPT-6 Astra (High) OpenAI	53.1%
16	GPT-6.1 Sol (Max) OpenAI	52.9%
17	GPT-6 Astra (Medium) OpenAI	52.7%
18	GPT-6.1 Sol (Xhigh) OpenAI	52.6%
19	Kimi K3.1 Pro Kimi	52.5%
20	GPT-6.1 Sol (High) OpenAI	51.4%
21	Muse Wonder (Max) Meta	50.1%
22	Claude Sonnet 5.5 (Xhigh, Default Fallback) Anthropic	50.0%
23	GPT-6.1 Sol (Medium) OpenAI	49.9%
24	MiMo-V2.6-Pro Xiaomi	49.4%
25	GPT-6 Astra (Low) OpenAI	49.2%
26	Claude Opus 5.5 (Low, Default Fallback) Anthropic	48.3%
27	MiniMax-M3.1-Flash MiniMax	48.0%
28	Gemini 3.8 Flash (High) Google	47.8%
29	GPT-6.1 Sol (Low) OpenAI	47.4%
30	Gemini 3.1 Pro Preview Google	47.0%
31	Step 5 Preview StepFun	46.5%
32	GPT-6.1 Terra (Max) OpenAI	46.2%
33	Claude Sonnet 5.5 (High, Default Fallback) Anthropic	45.8%
34	Claude Haiku 5.5 (Max) Anthropic	44.4%
35	Inkling 2 Small Thinking Machines	44.1%
36	Muse Spark Open (Xhigh) Meta	43.9%
37	GPT-6.1 Terra (Xhigh) OpenAI	43.7%
38	JT-4.1 Flash 236B A21B (Reasoning) China Mobile	43.6%
39	Ling 3.1 Flash 1019 InclusionAI	43.5%
40	DeepSeek V4.1 Flash 1017 (Max) DeepSeek	43.3%
41	GPT-6.1 Terra (High) OpenAI	43.3%
42	Grok 4.7 (Xhigh) SpaceXAI	43.1%
43	Qwen3.8 Max (0902) Alibaba	43.1%
44	Claude Haiku 5.5 (Xhigh) Anthropic	42.7%
45	GPT-5.3 Codex (Xhigh) OpenAI	42.5%
46	Qwen3.8 2.4T A95B Alibaba	42.4%
47	Grok 4.7 (High) SpaceXAI	42.3%
48	GLM-5.3 (Max) Z AI	42.3%
49	Gemini 3.8 Flash (Medium) Google	42.1%
50	DeepSeek V4 Pro 0813 (Max) DeepSeek	41.0%