CritPt

A benchmark designed to test LLMs on research-level physics reasoning tasks, featuring 71 composite research challenges.

Models: 265 current, 600 in total
Median: 0.3%
Range: 0.0%–35.7%

Score distribution

Lowest: 0.0%
Median: 0.3%
Highest: 35.7%

Leaderboard

Include deprecated models

Rank	Model	Score
1	Claude Fable 5.2 (Xhigh, Default Fallback) Anthropic	35.7%
2	Claude Fable 5.2 (Max, Default Fallback) Anthropic	35.4%
3	Claude Fable 5.2 (High, Default Fallback) Anthropic	32.9%
4	Muse Wonder Full (Max) Meta	32.9%
5	Claude Opus 5.5 (Max, Default Fallback) Anthropic	31.7%
6	Claude Opus 5.5 (Xhigh, Default Fallback) Anthropic	31.7%
7	GPT-6.1 Sol (Max) OpenAI	31.7%
8	GPT-6.1 Sol (Xhigh) OpenAI	31.7%
9	GPT-6 Astra (Max) OpenAI	31.7%
10	Claude Sonnet 5.5 (Max, Default Fallback) Anthropic	31.4%
11	GPT-6 Astra (Xhigh) OpenAI	31.4%
12	Claude Sonnet 5.5 (Xhigh, Default Fallback) Anthropic	31.1%
13	Claude Fable 5.2 (Medium, Default Fallback) Anthropic	30.9%
14	Claude Opus 5.5 (High, Default Fallback) Anthropic	30.9%
15	GPT-5.5 Pro (Xhigh) OpenAI	30.6%
16	GPT-6.1 Sol (High) OpenAI	30.0%
17	Claude Fable 5.2 (Low, Default Fallback) Anthropic	29.4%
18	GPT-6 Astra (Medium) OpenAI	29.1%
19	GPT-6 Astra (High) OpenAI	28.9%
20	Muse Wonder (Max) Meta	28.3%
21	GPT-6.1 Terra (Max) OpenAI	28.0%
22	Claude Opus 5.5 (Medium, Default Fallback) Anthropic	27.7%
23	GPT-6.1 Sol (Medium) OpenAI	27.7%
24	Gemini 4 Argon (High) Google	27.1%
25	GPT-6.1 Terra (Xhigh) OpenAI	26.6%
26	MiMo-V2.6-Pro Xiaomi	26.6%
27	GPT-6 Astra (Low) OpenAI	26.3%
28	Ling 3.1 Flash 1019 InclusionAI	26.0%
29	Gemini 3 Deep Think Google	25.7%
30	GPT-6.1 Terra (High) OpenAI	25.7%
31	GPT-6.1 Sol (Low) OpenAI	24.9%
32	Claude Sonnet 5.5 (High, Default Fallback) Anthropic	24.6%
33	Kimi K3.1 Pro Kimi	24.6%
34	DeepSeek V4.1 Flash 1017 (Max) DeepSeek	24.0%
35	Claude Haiku 5.5 (Xhigh) Anthropic	22.6%
36	MiniMax-M3.1-Flash MiniMax	22.3%
37	GPT-6.1 Terra (Medium) OpenAI	21.4%
38	Step 5 Preview StepFun	20.9%
39	Qwen3.8 2.4T A95B Alibaba	20.0%
40	GPT-6 Luna (Max) OpenAI	19.4%
41	GLM-5.3 (Max) Z AI	19.1%
42	Claude Haiku 5.5 (Max) Anthropic	18.9%
43	Claude Haiku 5.5 (High) Anthropic	18.6%
44	Gemini 3.8 Flash (High) Google	18.3%
45	DeepSeek V4 Pro 0813 (Max) DeepSeek	18.0%
46	Grok 4.7 (High) SpaceXAI	18.0%
47	Ling 3.1 Flash InclusionAI	18.0%
48	Claude Opus 5.5 (Low, Default Fallback) Anthropic	17.7%
49	Gemini 3.1 Pro Preview Google	17.7%
50	Grok 4.7 (Xhigh) SpaceXAI	17.7%