Skip to main content
BenchLM
Data

Critical Physics Tasks (CritPt)

Data verified

An Artificial Analysis metric for research-level physics reasoning.

Top models on CritPt — November 13, 2026

As of November 13, 2026, Claude Fable 5.2 leads the CritPt leaderboard with 35.4% , followed by Muse Wonder Full (32.9%) and GPT-5.6 Sol (32.3%, as reported by a third party).

1Closed

Claude Fable 5.2

Anthropic

Independent run: Artificial Analysis
35.4%
Overall 93.65Context 1M
2Closed

Muse Wonder Full

Meta

Independent run: Artificial Analysis
32.9%
Overall 72.73Context 1M
No rank on this tableClosed

GPT-5.6 Sol

OpenAI

Reported by a third party
32.3%
Overall 72.48Context 1.05M

233 modelsReasoning25% of Reasoning reference weightCurrentUpdated November 13, 2026

Leaderboard (233 models)

Of 233 results: 177 independent, 36 carried from an upstream page, 20 lab-reported.

Ranks on this table count the 177 benchmark-run and independent results. A lab-reported result, or one where who ran it is not recorded, keeps its place by score, without a rank.

Lab-reported
20
Independent
177
Benchmark-run
0
Carried from upstream
36
Who ran it

Showing all 233 results, in score order.

CritPt leaderboard
RankModel / configurationScoreParameters (B)Open / closed
1Claude Fable 5.2AnthropicIndependent run: Artificial Analysis
35.4%
Not reportedClosed
2Muse Wonder FullMetaIndependent run: Artificial Analysis
32.9%
Not reportedClosed
No rank on this tableGPT-5.6 SolOpenAIReported by a third party
32.3%
Not reportedClosed
3GPT-6 AstraOpenAIIndependent run: Artificial Analysis
31.7%
Not reportedClosed
4Claude Opus 5.5AnthropicIndependent run: Artificial Analysis
31.7%
Not reportedClosed
5GPT-6.1 SolOpenAIIndependent run: Artificial Analysis
31.7%
Not reportedClosed
6Claude Sonnet 5.5AnthropicIndependent run: Artificial Analysis
31.4%
Not reportedClosed
7GPT-6 SolOpenAIIndependent run: Artificial Analysis
30.9%
Not reportedClosed
No rank on this tableGPT-5.5 ProOpenAIReported by a third party
30.6%
Not reportedClosed
No rank on this tableGPT-5.6 TerraOpenAIReported by a third party
30.0%
Not reportedClosed
8GPT-5.4 ProOpenAIIndependent run: Artificial Analysis
30.0%
Not reportedClosed
9Claude Fable 5.1AnthropicIndependent run: Artificial Analysis
29.7%
Not reportedClosed
10Claude Opus 5AnthropicIndependent run: Artificial Analysis
29.1%
Not reportedClosed
11Claude Fable 5AnthropicIndependent run: Artificial Analysis
28.6%
Not reportedClosed
12Muse WonderMetaIndependent run: Artificial Analysis
28.3%
Not reportedClosed
13GPT-6.1 TerraOpenAIIndependent run: Artificial Analysis
28.0%
Not reportedClosed
14GPT-5.5OpenAIIndependent run: Artificial Analysis
27.1%
Not reportedClosed
15Gemini 4 ArgonGoogleIndependent run: Artificial Analysis
27.1%
Not reportedClosed
16MiMo-V2.6-ProXiaomiIndependent run: Artificial Analysis
26.6%
Not reportedOpen
No rank on this tableGemini 3 Pro Deep ThinkGoogleCarried from an upstream page
25.7%
Not reportedClosed
17Muse Spark 1.3MetaIndependent run: Artificial Analysis
24.9%
Not reportedClosed
18Kimi K3.1 ProMoonshot AIIndependent run: Artificial Analysis
24.6%
Not reportedOpen
19Kimi K3Moonshot AIIndependent run: Artificial Analysis
23.4%
Not reportedOpen
20GPT-5.4OpenAIIndependent run: Artificial Analysis
23.4%
Not reportedClosed
21MiniMax M3.1 FlashMiniMaxIndependent run: Artificial Analysis
22.3%
Not reportedClosed
22Claude Opus 4.8AnthropicIndependent run: Artificial Analysis
20.9%
Not reportedClosed
No rank on this tableGLM-5.2Z.AIReported by Z.AI
20.9%
Not reportedOpen
No rank on this tableStep 5 PreviewStepFunReported by StepFun
20.9%
Not reportedOpen
No rank on this tableGPT-5.6 LunaOpenAIReported by a third party
20.6%
Not reportedClosed
No rank on this tableSage-A1.5InternScienceReported by InternScience
20.0%
Not reportedOpen
23GPT-6 LunaOpenAIIndependent run: Artificial Analysis
19.4%
Not reportedClosed
24GLM-5.3Z.AIIndependent run: Artificial Analysis
19.1%
Not reportedOpen
25Claude Haiku 5.5AnthropicIndependent run: Artificial Analysis
18.9%
Not reportedClosed
26Gemini 3.8 FlashGoogleIndependent run: Artificial Analysis
18.3%
Not reportedClosed
No rank on this tableDeepSeek V4 Pro 0813DeepSeekCarried from an upstream page
18.0%
Not reportedOpen
27Ling 3.1 FlashInclusionAIIndependent run: Artificial Analysis
18.0%
Not reportedOpen
No rank on this tableGemini 3.1 ProGoogleCarried from an upstream page
17.7%
Not reportedClosed
28Muse Spark 1.2MetaIndependent run: Artificial Analysis
17.7%
Not reportedClosed
29Grok 4.7xAIIndependent run: Artificial Analysis
17.7%
Not reportedClosed
No rank on this tableQwen3.8 Max PreviewAlibabaCarried from an upstream page
17.7%
Not reportedClosed
No rank on this tableBeam 1.0 501B-A23B FP8Reflection AIReported by Reflection AI
17.4%
Not reportedOpen
30Grok 4.6xAIIndependent run: Artificial Analysis
17.1%
Not reportedClosed
31Claude Sonnet 5AnthropicIndependent run: Artificial Analysis
16.9%
Not reportedClosed
32GPT-5.3 CodexOpenAIIndependent run: Artificial Analysis
16.9%
Not reportedClosed
No rank on this tableHy4 previewTencentReported by Tencent
16.9%
Not reportedOpen
No rank on this tableGPT-5.3-Codex-SparkOpenAICarried from an upstream page
16.9%
Not reportedClosed
No rank on this tableDeepSeek V4 Flash 0731DeepSeekCarried from an upstream page
16.6%
Not reportedOpen
No rank on this tableBeam 1.0 501B-A23BReflection AIReported by Reflection AI
16.3%
Not reportedOpen
No rank on this tableBeam 1.0 501B-A23B NVFP4Reflection AIReported by Reflection AI
15.4%
Not reportedOpen
No rank on this tableGrok 4.5xAIReported by a third party
15.4%
Not reportedClosed
33Muse Spark 1.1MetaIndependent run: Artificial Analysis
15.1%
Not reportedClosed
34Muse Spark OpenMetaIndependent run: Artificial Analysis
14.9%
Not reportedOpen
35Gemini 3.7 FlashGoogleIndependent run: Artificial Analysis
14.3%
Not reportedClosed
36DeepSeek V4.1 Flash 1017DeepSeekIndependent run: Artificial Analysis
14.3%
Not reportedOpen
No rank on this tableQwen3.7 MaxAlibabaReported by Alibaba
13.4%
Not reportedClosed
No rank on this tableGemini 3.5 FlashGoogleReported by a third party
13.1%
Not reportedClosed
No rank on this tableClaude Opus 4.6 (Adaptive)AnthropicCarried from an upstream page
12.6%
Not reportedClosed
No rank on this tableClaude Opus 4.7 (Adaptive)AnthropicCarried from an upstream page
12.0%
Not reportedClosed
37MiMo-V2.6-FlashXiaomiIndependent run: Artificial Analysis
12.0%
Not reportedOpen
38GPT-5.2OpenAIIndependent run: Artificial Analysis
11.6%
Not reportedClosed
39Muse SparkMetaIndependent run: Artificial Analysis
11.3%
Not reportedClosed
40Qwen3.8-Flash-NextAlibabaIndependent run: Artificial Analysis
11.1%
Not reportedOpen
No rank on this tableNemotron 3.5 Super VL 120B A12BNVIDIAReported by NVIDIA
10.9%
Not reportedOpen
41Gemini 3.6 FlashGoogleIndependent run: Artificial Analysis
10.6%
Not reportedClosed
No rank on this tableNemotron 3.5 Super VL 120B A12B NVFP4NVIDIACarried from an upstream page
10.6%
Not reportedOpen
42Kimi K2.7 CodeMoonshot AIIndependent run: Artificial Analysis
10.0%
Not reportedOpen
43GPT-5.4 miniOpenAIIndependent run: Artificial Analysis
10.0%
Not reportedClosed
44Mistral Large 4MistralIndependent run: Artificial Analysis
10.0%
Not reportedOpen
45DeepSeek V4 Pro (High)DeepSeekIndependent run: Artificial Analysis
10.0%
Not reportedOpen
46Quasar 438BMultiverse ComputingIndependent run: Artificial Analysis
9.4%
Not reportedClosed
47GPT-5.4 nanoOpenAIIndependent run: Artificial Analysis
9.3%
Not reportedClosed
No rank on this tableQwen3.7 PlusAlibabaReported by Alibaba
9.1%
Not reportedClosed
48Gemini 3 ProGoogleIndependent run: Artificial Analysis
9.1%
Not reportedClosed
49A.X K2SK TelecomIndependent run: Artificial Analysis
8.9%
Not reportedOpen
50GPT-5.2-CodexOpenAIIndependent run: Artificial Analysis
8.7%
Not reportedClosed
No rank on this tableInkling-SmallThinking Machines LabReported by Thinking Machines Lab
8.3%
Not reportedOpen
No rank on this tableKimi K2.6Moonshot AICarried from an upstream page
8.0%
Not reportedOpen
No rank on this tableGrok 4.3xAIReported by a third party
8.0%
Not reportedClosed
No rank on this tableSolar Pro 4UpstageCarried from an upstream page
6.3%
Not reportedClosed
No rank on this tableGPT-5 (high)OpenAICarried from an upstream page
5.7%
Not reportedClosed
51GPT-5.1-CodexOpenAIIndependent run: Artificial Analysis
5.7%
Not reportedClosed
No rank on this tableGPT-5.1-Codex-MaxOpenAICarried from an upstream page
5.7%
Not reportedClosed
52InklingThinking Machines LabIndependent run: Artificial Analysis
5.4%
Not reportedOpen
53Qwen3.8-27BAlibabaIndependent run: Artificial Analysis
5.4%
Not reportedOpen
No rank on this tableClaude Opus 4.7AnthropicCarried from an upstream page
5.1%
Not reportedClosed
54GPT-5.1OpenAIIndependent run: Artificial Analysis
4.9%
Not reportedClosed
55Hy3TencentIndependent run: Artificial Analysis
4.9%
Not reportedOpen
No rank on this tableHy3 PreviewTencentReported by a third party
4.9%
Not reportedOpen
56GLM-5.1Z.AIIndependent run: Artificial Analysis
4.6%
Not reportedOpen
57Apodex 1.1ApodexIndependent run: Artificial Analysis
4.6%
Not reportedClosed
58Claude Opus 4.5 ThinkingAnthropicIndependent run: Artificial Analysis
4.6%
Not reportedClosed
No rank on this tableApodex 1.1 MiniApodexCarried from an upstream page
4.6%
Not reportedOpen
59Nemotron 3.5 Bolt 74B-A9BNVIDIAIndependent run: Artificial Analysis
4.3%
Not reportedOpen
60MiMo-V2.5-ProXiaomiIndependent run: Artificial Analysis
4.0%
Not reportedClosed
61MiniMax M3MiniMaxIndependent run: Artificial Analysis
3.7%
Not reportedOpen
No rank on this tableQwen 3.6 Max (preview)AlibabaCarried from an upstream page
3.7%
Not reportedClosed
No rank on this tableKimi K2.5 (Reasoning)Moonshot AICarried from an upstream page
3.1%
Not reportedOpen
62Kimi K2.5Moonshot AIIndependent run: Artificial Analysis
3.1%
Not reportedOpen
No rank on this tableNemotron 3 UltraNVIDIAReported by NVIDIA
3.1%
Not reportedOpen
63Qwen3.6 PlusAlibabaIndependent run: Artificial Analysis
2.9%
Not reportedClosed
64Grok 4.1 Fast (Reasoning)xAIIndependent run: Artificial Analysis
2.9%
Not reportedClosed
65Grok 4 Fast (Reasoning)xAIIndependent run: Artificial Analysis
2.9%
Not reportedClosed
66Claude Opus 4.6AnthropicIndependent run: Artificial Analysis
2.8%
Not reportedClosed
67Muse Glimmer 30BMetaIndependent run: Artificial Analysis
2.6%
Not reportedOpen
68Gemini 2.5 ProGoogleIndependent run: Artificial Analysis
2.6%
Not reportedClosed
69Step 3.7 FlashStepFunIndependent run: Artificial Analysis
2.3%
Not reportedOpen
70Step 3.5 FlashStepFunIndependent run: Artificial Analysis
2.3%
Not reportedOpen
71DeepSeek V3.1 (Reasoning)DeepSeekIndependent run: Artificial Analysis
2.0%
Not reportedOpen
72GLM-5Z.AIIndependent run: Artificial Analysis
2.0%
Not reportedOpen
73Ling 3.0 Flash VLInclusionAIIndependent run: Artificial Analysis
2.0%
Not reportedOpen
74Grok 4xAIIndependent run: Artificial Analysis
2.0%
Not reportedClosed
75Ling 3.0 FlashInclusionAIIndependent run: Artificial Analysis
1.7%
Not reportedOpen
No rank on this tableQwen3.5 397BAlibabaCarried from an upstream page
1.7%
Not reportedOpen
76GLM-4.7Z.AIIndependent run: Artificial Analysis
1.7%
Not reportedOpen
No rank on this tableQwen3.5 397B (Reasoning)AlibabaCarried from an upstream page
1.7%
Not reportedOpen
77Gemma 4 31BGoogleIndependent run: Artificial Analysis
1.4%
Not reportedOpen
78DeepSeek-R1DeepSeekIndependent run: Artificial Analysis
1.4%
Not reportedOpen
79Gemini 3 FlashGoogleIndependent run: Artificial Analysis
1.4%
Not reportedClosed
80Gemini 2.5 FlashGoogleIndependent run: Artificial Analysis
1.4%
Not reportedClosed
81GPT-OSS 20BOpenAIIndependent run: Artificial Analysis
1.4%
Not reportedOpen
82o3OpenAIIndependent run: Artificial Analysis
1.1%
Not reportedClosed
83Qwen3.6-27BAlibabaIndependent run: Artificial Analysis
1.1%
Not reportedOpen
84GPT-OSS 120BOpenAIIndependent run: Artificial Analysis
1.1%
Not reportedOpen
85MiMo-V2-OmniXiaomiIndependent run: Artificial Analysis
1.1%
Not reportedClosed
86Claude 4 SonnetAnthropicIndependent run: Artificial Analysis
1.1%
Not reportedClosed
87K-ExaoneLG AI ResearchIndependent run: Artificial Analysis
1.1%
Not reportedClosed
88MiniMax M2.5MiniMaxIndependent run: Artificial Analysis
1.1%
Not reportedClosed
89Claude Sonnet 4.6AnthropicIndependent run: Artificial Analysis
0.9%
Not reportedClosed
90Qwen3.5-27BAlibabaIndependent run: Artificial Analysis
0.9%
Not reportedOpen
91Qwen3.5-35B-A3BAlibabaIndependent run: Artificial Analysis
0.9%
Not reportedOpen
92DeepSeek V3.2DeepSeekIndependent run: Artificial Analysis
0.9%
Not reportedOpen
93Claude 3.7 Sonnet ThinkingAnthropicIndependent run: Artificial Analysis
0.9%
Not reportedClosed
94Trinity-Large-ThinkingArcee AIIndependent run: Artificial Analysis
0.9%
Not reportedOpen
No rank on this tableTrinity-Large-PreviewArcee AICarried from an upstream page
0.9%
Not reportedOpen
No rank on this tableK-EXAONE 2.0LG AI ResearchCarried from an upstream page
0.9%
Not reportedOpen
95Mercury 2InceptionIndependent run: Artificial Analysis
0.8%
Not reportedClosed
96Qwen3.5-122B-A10BAlibabaIndependent run: Artificial Analysis
0.6%
Not reportedOpen
97MiniMax M2.7MiniMaxIndependent run: Artificial Analysis
0.6%
Not reportedOpen
98Gemma 4 E4BGoogleIndependent run: Artificial Analysis
0.6%
Not reportedOpen
99GLM-5V-TurboZ.AIIndependent run: Artificial Analysis
0.6%
Not reportedClosed
100o4-miniOpenAIIndependent run: Artificial Analysis
0.6%
Not reportedClosed
101Claude Opus 4.5AnthropicIndependent run: Artificial Analysis
0.3%
Not reportedClosed
102Qwen3.6-35B-A3BAlibabaIndependent run: Artificial Analysis
0.3%
Not reportedOpen
103GLM-5-TurboZ.AIIndependent run: Artificial Analysis
0.3%
Not reportedClosed
104MiMo-V2-ProXiaomiIndependent run: Artificial Analysis
0.3%
Not reportedClosed
105o1OpenAIIndependent run: Artificial Analysis
0.3%
Not reportedClosed
106Mistral Small 4MistralIndependent run: Artificial Analysis
0.3%
Not reportedOpen
107Granite 4.2 30BIBMIndependent run: Artificial Analysis
0.3%
Not reportedOpen
108Command A+CohereIndependent run: Artificial Analysis
0.3%
Not reportedOpen
109Granite 4.2 8BIBMIndependent run: Artificial Analysis
0.3%
Not reportedOpen
110Magistral MediumMistralIndependent run: Artificial Analysis
0.3%
Not reportedClosed
111Sarvam 30BSarvamIndependent run: Artificial Analysis
0.3%
Not reportedOpen
112MiniCPM5-2BOpenBMBIndependent run: Artificial Analysis
0.3%
Not reportedOpen
No rank on this tableNorth Mini CodeCohereCarried from an upstream page
0.3%
Not reportedOpen
No rank on this tableMistral Small 4 (Reasoning)MistralCarried from an upstream page
0.3%
Not reportedOpen
113GLM-4.7-FlashZ.AIIndependent run: Artificial Analysis
0.3%
Not reportedOpen
114Gemma 4 12BGoogleIndependent run: Artificial Analysis
0.0%
Not reportedOpen
115Gemma 4 26B A4BGoogleIndependent run: Artificial Analysis
0.0%
Not reportedOpen
116DeepSeek V3 0324DeepSeekIndependent run: Artificial Analysis
0.0%
Not reportedOpen
117Kimi K2Moonshot AIIndependent run: Artificial Analysis
0.0%
Not reportedClosed
118GPT-4.1OpenAIIndependent run: Artificial Analysis
0.0%
Not reportedClosed
119Nemotron 3 Nano Omni 30B A3BNVIDIAIndependent run: Artificial Analysis
0.0%
Not reportedOpen
120GPT-4oOpenAIIndependent run: Artificial Analysis
0.0%
Not reportedClosed
121Phi-4MicrosoftIndependent run: Artificial Analysis
0.0%
Not reportedOpen
122Gemma 4 E2BGoogleIndependent run: Artificial Analysis
0.0%
Not reportedOpen
123GPT-5 (medium)OpenAIIndependent run: Artificial Analysis
0.0%
Not reportedClosed
124Claude 4.1 Opus ThinkingAnthropicIndependent run: Artificial Analysis
0.0%
Not reportedClosed
125Exaone 4.0 32BLG AI ResearchIndependent run: Artificial Analysis
0.0%
Not reportedOpen
126Mistral Medium 3.5 128BMistralIndependent run: Artificial Analysis
0.0%
Not reportedOpen
127Gemini 3.5 Flash-LiteGoogleIndependent run: Artificial Analysis
0.0%
Not reportedClosed
128Qwen3.5-4BAlibabaIndependent run: Artificial Analysis
0.0%
Not reportedOpen
129Ling 3.0 TinyInclusionAIIndependent run: Artificial Analysis
0.0%
Not reportedOpen
No rank on this tableNemotron Ultra 253BNVIDIACarried from an upstream page
0.0%
Not reportedOpen
130Qwen3 MaxAlibabaIndependent run: Artificial Analysis
0.0%
Not reportedClosed
No rank on this tableLlama 3.3 70BMetaCarried from an upstream page
0.0%
Not reportedOpen
131Celeris-1CelerisIndependent run: Artificial Analysis
0.0%
Not reportedClosed
No rank on this tableLlama 3.1 70BMetaCarried from an upstream page
0.0%
Not reportedOpen
132Llama 4 MaverickMetaIndependent run: Artificial Analysis
0.0%
Not reportedOpen
133GLM-4.6Z.AIIndependent run: Artificial Analysis
0.0%
Not reportedOpen
134Mercury 2.5InceptionIndependent run: Artificial Analysis
0.0%
Not reportedClosed
135Claude 3.7 SonnetAnthropicIndependent run: Artificial Analysis
0.0%
Not reportedClosed
136Grok Code Fast 1xAIIndependent run: Artificial Analysis
0.0%
Not reportedClosed
137MiMo-V2-FlashXiaomiIndependent run: Artificial Analysis
0.0%
Not reportedOpen
138DeepSeek V3.1DeepSeekIndependent run: Artificial Analysis
0.0%
Not reportedOpen
139Solar Pro 3UpstageIndependent run: Artificial Analysis
0.0%
Not reportedClosed
140GLM-4.5-AirZ.AIIndependent run: Artificial Analysis
0.0%
Not reportedClosed
No rank on this tableNemotron 3 Nano 30BNVIDIACarried from an upstream page
0.0%
Not reportedOpen
141Llama 4 ScoutMetaIndependent run: Artificial Analysis
0.0%
Not reportedOpen
No rank on this tablePhi-4-reasoningMicrosoftCarried from an upstream page
0.0%
Not reportedOpen
142Mistral Large 3MistralIndependent run: Artificial Analysis
0.0%
Not reportedOpen
143Mistral Medium 3MistralIndependent run: Artificial Analysis
0.0%
Not reportedClosed
144Mistral Large 2MistralIndependent run: Artificial Analysis
0.0%
Not reportedClosed
145Granite 4.2 3BIBMIndependent run: Artificial Analysis
0.0%
Not reportedOpen
146GPT-4.1 miniOpenAIIndependent run: Artificial Analysis
0.0%
Not reportedClosed
147Grok 4.1 FastxAIIndependent run: Artificial Analysis
0.0%
Not reportedClosed
148Claude 3 HaikuAnthropicIndependent run: Artificial Analysis
0.0%
Not reportedClosed
149Ling 2.6 FlashInclusionAIIndependent run: Artificial Analysis
0.0%
Not reportedOpen
150Magistral SmallMistralIndependent run: Artificial Analysis
0.0%
Not reportedOpen
151LFM2.5-2.6BLiquidAIIndependent run: Artificial Analysis
0.0%
Not reportedOpen
152Qwen3-Omni-30B-A3B-InstructAlibabaIndependent run: Artificial Analysis
0.0%
Not reportedOpen
153Gemini 2.0 FlashGoogleIndependent run: Artificial Analysis
0.0%
Not reportedClosed
No rank on this tableNemotron 3.5 Lightning 30B A3B NVFP4NVIDIACarried from an upstream page
0.0%
Not reportedOpen
154Mistral Small 3.1MistralIndependent run: Artificial Analysis
0.0%
Not reportedOpen
155Nova ProAmazonIndependent run: Artificial Analysis
0.0%
Not reportedClosed
156Gemini 2.5 Flash-LiteGoogleIndependent run: Artificial Analysis
0.0%
Not reportedClosed
157DeepSeek V3DeepSeekIndependent run: Artificial Analysis
0.0%
Not reportedOpen
158Granite-4.0-H-1BIBMIndependent run: Artificial Analysis
0.0%
Not reportedOpen
159Gemma 3 27BGoogleIndependent run: Artificial Analysis
0.0%
Not reportedOpen
160Solar Pro 2UpstageIndependent run: Artificial Analysis
0.0%
Not reportedClosed
161Exaone 4.0 1.2BLG AI ResearchIndependent run: Artificial Analysis
0.0%
Not reportedOpen
162Devstral SmallMistralIndependent run: Artificial Analysis
0.0%
Not reportedOpen
163Claude 3.5 HaikuAnthropicIndependent run: Artificial Analysis
0.0%
Not reportedClosed
164Sarvam 105BSarvamIndependent run: Artificial Analysis
0.0%
Not reportedOpen
165LFM2.5-8B-A1BLiquidAIIndependent run: Artificial Analysis
0.0%
Not reportedOpen
166Granite-4.0-H-350MIBMIndependent run: Artificial Analysis
0.0%
Not reportedOpen
167Granite-4.0-350MIBMIndependent run: Artificial Analysis
0.0%
Not reportedOpen
168GPT-4.1 nanoOpenAIIndependent run: Artificial Analysis
0.0%
Not reportedClosed
No rank on this tableLlama 3.1 405BMetaCarried from an upstream page
0.0%
Not reportedOpen
No rank on this tableLFM2.5-VL-1.6B-ExtractLiquidAICarried from an upstream page
0.0%
Not reportedOpen
No rank on this tableQwen3-Omni-30B-A3B-ThinkingAlibabaCarried from an upstream page
0.0%
Not reportedOpen
No rank on this tableUltravox v0.6 Llama 3.3 70BFixie AICarried from an upstream page
0.0%
Not reportedOpen
169GPT-5 miniOpenAIIndependent run: Artificial Analysis
0.0%
Not reportedClosed
No rank on this tableMinistral 3 14B (Reasoning)MistralCarried from an upstream page
0.0%
Not reportedOpen
170Ministral 3 14BMistralIndependent run: Artificial Analysis
0.0%
Not reportedOpen
171GPT-5 nanoOpenAIIndependent run: Artificial Analysis
0.0%
Not reportedClosed
172MiniMax M1 80kMiniMaxIndependent run: Artificial Analysis
0.0%
Not reportedClosed
173LFM2-24B-A2BLiquidAIIndependent run: Artificial Analysis
0.0%
Not reportedClosed
No rank on this tableMinistral 3 8B (Reasoning)MistralCarried from an upstream page
0.0%
Not reportedOpen
174LFM2.5-1.2B-ThinkingLiquidAIIndependent run: Artificial Analysis
0.0%
Not reportedClosed
175Ministral 3 8BMistralIndependent run: Artificial Analysis
0.0%
Not reportedOpen
No rank on this tableMinistral 3 3B (Reasoning)MistralCarried from an upstream page
0.0%
Not reportedOpen
176LFM2.5-1.2B-InstructLiquidAIIndependent run: Artificial Analysis
0.0%
Not reportedClosed
177Ministral 3 3BMistralIndependent run: Artificial Analysis
0.0%
Not reportedOpen

A dash in the rank column marks a lab-reported result or a result carried from an upstream page, where who ran it is not recorded. It keeps its place by score, without a rank on this table.

  • Lab-reported
  • Benchmark-run
  • Independent
What the labels mean

According to BenchLM.ai, Claude Fable 5.2 leads the CritPt benchmark with a score of 35.4%, followed by Muse Wonder Full (32.9%) and GPT-5.6 Sol (32.3%, as reported by a third party). The scores show moderate spread, with meaningful differences between the top tier and mid-tier models.

233 models have been evaluated on CritPt. The benchmark falls in the Reasoning category. BenchAlign v5.8 gives CritPt 25% of the Reasoning reference weight, so it moves the Reasoning leaderboard and the overall ranking. Reference weights are relative weights in the calibrated model, not fixed shares of a score.

About CritPt

Year

2026

Tasks

Research-level physics questions

Format

Accuracy

Difficulty

Research-level physics reasoning

BenchLM stores CritPt results from Artificial Analysis' independently evaluated leaderboard.

CritPt Benchmark Leaderboard

Freshness and provenance

Version

CritPt 2026

Refresh cadence

Quarterly

Staleness state

Current

Question availability

Public benchmark set

Current

BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.

Questions

What does CritPt measure?

An Artificial Analysis metric for research-level physics reasoning.

Which model scores highest on CritPt?

Claude Fable 5.2 by Anthropic currently leads with a score of 35.4% on CritPt.

How many models are evaluated on CritPt?

233 AI models have published results on CritPt in the BenchLM catalog.

Compare top models on CritPt

Claude Fable 5.2 vs Muse Wonder FullMuse Wonder Full vs GPT-5.6 SolGPT-5.6 Sol vs GPT-6 AstraGPT-6 Astra vs Claude Opus 5.5
Last updated: November 13, 2026 · BenchLM version CritPt 2026

Know when it’s worth switching models

The model to choose, the cheaper alternative, and the release we would wait on.

Read a sample issue

Join 7,000+ subscribers.

One email each week. Unsubscribe anytime.