CritPt

一项基准，用于测试 LLM 在研究级物理推理任务上的表现，包含 71 个复合研究挑战。

模型: 265 当前，600 总共
中位数: 0.3%
范围: 0.0%–35.7%

分数分布

最低: 0.0%
中位数: 0.3%
最高: 35.7%

排行榜

包括已弃用的模型

数据来源：Artificial Analysis

排名	比较	模型	评分
101		Ling-3.0-flash-VL InclusionAI	2.0%
102		Qwen3.5 397B A17B (Reasoning) Alibaba	1.7%
103		ERNIE 5.0 Thinking Preview Baidu	1.4%
104		Gemma 4 31B (Reasoning) Google	1.4%
105		gpt-oss-20b (High) OpenAI	1.4%
106		Solar Mini 4 Upstage	1.4%
107		GPT-6 Luna (Non-reasoning) OpenAI	1.1%
108		gpt-oss-120b (High) OpenAI	1.1%
109		o3 OpenAI	1.1%
110		HyperNova 60B 2605 (High, Based on gpt-oss-120b) Multiverse Computing	0.9%
111		K-EXAONE 2.0 LG AI Research	0.9%
112		NVIDIA Nemotron 3 Nano 30B A3B (Reasoning) NVIDIA	0.9%
113		Qwen3.5 122B A10B (Non-reasoning) Alibaba	0.9%
114		Qwen3.5 397B A17B (Non-reasoning) Alibaba	0.9%
115		Trinity Large Thinking Arcee AI	0.9%
116		Gemma 4 E4B (Reasoning) Google	0.6%
117		Nemotron Cascade 2 30B A3B NVIDIA	0.6%
118		Qwen3.5 122B A10B (Reasoning) Alibaba	0.6%
119		Qwen3.5 9B (Non-reasoning) Alibaba	0.6%
120		Qwen3.5 Omni Plus Alibaba	0.6%
121		Apriel-v1.6-15B-Thinker ServiceNow	0.3%
122		Command A+ Cohere	0.3%
123		DeepSeek V4.1 Flash 1017 (Non-reasoning) DeepSeek	0.3%
124		DiffusionGemma 26B A4B Google	0.3%
125		Doubao Seed Code ByteDance Seed	0.3%
126		EXAONE 4.5 33B (Reasoning) LG AI Research	0.3%
127		Falcon-H1R-7B TII UAE	0.3%
128		G9v3-39A5B AI9Stars	0.3%
129		Gemma 4 E4B (Non-reasoning) Google	0.3%
130		Granite 4.2 30B IBM	0.3%
131		Granite 4.2 8B IBM	0.3%
132		Hermes 4 - Llama-3.1 405B (Reasoning) Nous Research	0.3%
133		INTELLECT-3 (Based on GLM-4.5-Air) Prime Intellect	0.3%
134		K2 Horizon 7B Institute of Foundation Models	0.3%
135		Magistral Medium 1.2 Mistral	0.3%
136		Magistral Small 1.2 Mistral	0.3%
137		MiniCPM5-2B OpenBMB	0.3%
138		Mistral Small 4 (Reasoning) Mistral	0.3%
139		Mistral Small 4 (Non-reasoning) Mistral	0.3%
140		North Mini Code Cohere	0.3%
141		Nova 2.0 Lite (High) Amazon	0.3%
142		Qwen3.5 9B (Reasoning) Alibaba	0.3%
143		Qwen3.6 35B A3B (Reasoning) Alibaba	0.3%
144		Qwen3.8 27B (Non-reasoning) Alibaba	0.3%
145		Sarvam 30B (High) Sarvam	0.3%
146		Tri-21B-Think Trillion Labs	0.3%
147		Ring-flash-2.0 InclusionAI	0.3%
148		Apertus 70B Instruct Swiss AI Initiative	0.0%
149		Apertus 8B Instruct Swiss AI Initiative	0.0%
150		Celeris-1 Celeris	0.0%

101
Ling-3.0-flash-VL
InclusionAI
2.0%
102
Qwen3.5 397B A17B (Reasoning)