The AI tools index that doesn't waste your time.不浪费你时间的 AI 工具索引。
Reasoning · weight 15%推理 · 权重 15%

GPQA Diamond

official site官网 ↗

What it tests: Graduate-level multiple-choice science questions (biology, physics, chemistry) written so that experts with web access still find them hard. 'Diamond' is the highest-quality 198-question subset.

考什么:研究生级别的科学选择题(生物/物理/化学),专门设计成'专家带着搜索引擎也觉得难'。Diamond 是其中质量最高的 198 题子集。

Why it counts: One of the last knowledge benchmarks where frontier models still make mistakes — it separates genuine scientific reasoning from memorized answers.

为什么算数:少数还能让顶尖模型犯错的 benchmark 之一——区分真科学推理和背答案。头部模型约 80-96%,尚未触顶。

Maintainer维护方NYU / Anthropic 等联合
License许可学术免费,商用需授权
Reference point (prior-gen SOTA)参考点(上一代 SOTA)80
◈

Standings实时排名

59 models with sourced scores 个模型有溯源成绩
#Model模型Score成绩Index指数Dated日期Measured by测评方
1Gemini 3.1 Pro PreviewGoogle95.5100—3rd-party第三方source ↗×9
2Claude Opus 4.8Anthropic94.3992026-07-013rd-party第三方source ↗×4
3Claude Fable 5Anthropic94.1⚡max99—3rd-party第三方source ↗×8⚠±15.7
4GPT-5.6 SolOpenAI94.1⚡max99—3rd-party第三方source ↗×7
5GPT-5.6 Sol ProOpenAI94.1⚡high99—3rd-party第三方source ↗×2
6Claude Opus 5Anthropic93.798—3rd-party第三方source ↗×11⚠±5.2
7Kimi K3Moonshot AI93.5982026-07-183rd-party第三方source ↗×7
8GPT-5.5OpenAI93.5⚡high98—3rd-party第三方source ↗×7
9Gemini 3.6 FlashGoogle93.43982026-083rd-party第三方source ↗×5
10Grok 4.5xAI92.9972026-07-013rd-party第三方source ↗×6⚠±6.5
11Qwen3.8 Max PrimeQwen92.697—3rd-party第三方source ↗×4
12Qwen3.7 MaxQwen92.497—3rd-party第三方source ↗×6
13Gemini 3.5 FlashGoogle92.2⚡high97—3rd-party第三方source ↗×5
14GLM 5.2Z.ai91.2962026-07-183rd-party第三方source ↗×9⚠±3.8
15MiniMax M3MiniMax91.296—3rd-party第三方source ↗×5
16GPT-5.6 TerraOpenAI91.1⚡high95—3rd-party第三方source ↗×6⚠±4.21
17GPT-5.6 LunaOpenAI91.1⚡max95—3rd-party第三方source ↗×3
18Grok 4.20xAI9195—3rd-party第三方source ↗×4⚠±3.5
19DeepSeek V4 Flash 0731DeepSeek90.895—3rd-party第三方source ↗×2
20Hy3Tencent90.4952026-083rd-party第三方source ↗×2
21Muse Spark 1.2meta90.495—3rd-party第三方source ↗
22Hunyuan A13B InstructTencent90.495—3rd-party第三方source ↗×2⚠±7.7
23Qwen3.7 PlusQwen90.3952026-063rd-party第三方source ↗×4
24Qwen3.7 FlashQwen90.1594—3rd-party第三方source ↗×2
25DeepSeek V4 Pro 0423DeepSeek90.1942026-07-183rd-party第三方source ↗×9⚠±4.7
26Grok 4.3xAI90.1⚡high94—3rd-party第三方source ↗×4
27Muse Spark 1.1meta89.5942026-083rd-party第三方source ↗×2
28GPT-5.6 Luna ProOpenAI89.594—3rd-party第三方source ↗
29DeepSeek V4 Flash 0423DeepSeek87.492—3rd-party第三方source ↗×5⚠±3.9
30Nemotron 3 UltraNVIDIA8791—3rd-party第三方source ↗×2
31MiMo-V2.5-ProXiaomi8791—3rd-party第三方source ↗×2
32GLM 5.1Z.ai87⚡high91—3rd-party第三方source ↗×5⚠±3.7
33Gemini 3.1 Flash LiteGoogle86.991—3rd-party第三方source ↗×5
34MiniMax M2.7MiniMax86.6912026-03-013rd-party第三方source ↗×7⚠±5.4
35Seed-2.0-LiteByteDance Seed85.189—3rd-party第三方source ↗×2
36Kimi K2 ThinkingMoonshot AI84.5⚡high88—3rd-party第三方source ↗×7
37Gemini 3.5 Flash LiteGoogle8488—3rd-party第三方source ↗×3
38DeepSeek V3.2DeepSeek82.987—3rd-party第三方source ↗×5⚠±5.4
39gpt-oss-120bOpenAI80.1⚡high84—3rd-party第三方source ↗×3
40Seed-2.0-MiniByteDance Seed7983—3rd-party第三方source ↗×2
41LongCat 2.0Meituan7882—3rd-party第三方source ↗×2⚠±10.9
42Kimi K2.7 CodeMoonshot AI77.881—3rd-party第三方source ↗×5⚠±13.4
43Mercury 2Inception7781—3rd-party第三方source ↗×2
44Step 3.7 FlashStepFun76.680—3rd-party第三方source ↗×3⚠±4.3
45Mistral Medium 3.5Mistral AI74.878—3rd-party第三方source ↗×5
46Claude Sonnet 5Anthropic74.678—3rd-party第三方source ↗×6⚠±21.6
47Sonar ProPerplexity72.476—3rd-party第三方source ↗×2⚠±10.1
48Solar Pro 3Upstage72.476—3rd-party第三方source ↗
49Llama 4 MaverickMeta69.873—3rd-party第三方source ↗×8
50Claude Haiku 4.5Anthropic64.668—3rd-party第三方source ↗×5⚠±22.9
51Llama 4 ScoutMeta57.260—3rd-party第三方source ↗×6
52Nova Premier 1.0Amazon56.960—3rd-party第三方source ↗×4
53Phi 4Microsoft56.159—3rd-party第三方source ↗×3⚠±13.8
54Command ACohere50.853—3rd-party第三方source ↗×4⚠±26.1
55Nova Pro 1.0Amazon49.952—3rd-party第三方source ↗×2⚠±3
56Palmyra X5Writer47.249—vendor-reported厂商自报source ↗
57Mistral Small 4Mistral AI44.2462026-04-013rd-party第三方source ↗×2⚠±6.1
58Mistral LargeMistral AI43.94620263rd-party第三方source ↗×6⚠±24.1
59Command R+ (08-2024)Cohere31.133—3rd-party第三方source ↗×2

How scores become the Intelligence Score →成绩如何合成智能评分 →