Benchmarks / Reasoning

CritPt

CritPt: research-level physics reasoning problems; accuracy.

368models
–Open Index weight
–last updated

CritPt

Higher is better · best effort per model

Chart settings
368 of 368 selected

full table below

#ModelEffortScoreCost / taskHarness
1Muse Wonder Full (Max)max32.9%––
2GPT-5.6 Solmax32.3%$0.859–
3Opus 5.5max31.7%$2.19–
4GPT-6 Astramax31.7%$1.17–
5GPT-6.1 Solmax31.7%$0.242–
6Sonnet 5.5max31.4%$1.89–
7Fable 5.1xhigh31.1%$4.53–
8GPT-6 Solmax30.9%$0.324–
9GPT-5.5 Proxhigh30.6%––
10GPT-5.6 Terramax30.0%$0.608–
11GPT-5.4 Proxhigh30.0%––
12Opus 5max29.1%$2.88–
13Fable 5max28.6%$5.62–
14Muse Wondermax28.3%$1.49–
15Gemini 4high27.1%$1.58–
16GPT-5.5xhigh27.1%$1.61–
17MiMo V2.6 Pro–26.6%$0.060–
18Muse Spark 1.3xhigh26.0%$0.308–
19Gemini 3–25.7%––
20GPT-5.4xhigh23.4%––
21Kimi K3max23.4%$0.902–
22Haiku 5.5xhigh22.6%$0.065–
23Step 5 Preview–20.9%$0.158–
24Opus 4.8max20.9%$1.94–
25GLM 5.2max20.9%$0.482–
26GPT-5.6 Lunamax20.6%$0.063–
27Qwen 3.8 Maxmax20.0%$0.509–
28Qwen 3.8 2.4T-A95B–20.0%$0.483–
29Grok 4.6xhigh19.7%$0.311–
30GPT-6 Lunamax19.4%$0.027–
31GLM 5.3max19.1%$0.452–
32Gemini 3.8 Flashhigh18.3%$0.280–
33Grok 4.7high18.0%$0.681–
34Ling 3.1 Flash–18.0%$0.146–
35DeepSeek V4 Pro 0813max18.0%$0.357–
36Muse Spark 1.2xhigh17.7%$0.346–
37Gemini 3.1 Pro–17.7%$0.474–
38Sonnet 5max16.9%$2.31–
39GPT-5.3 Codexxhigh16.9%––
40DeepSeek V4 Flash 0731max16.6%$0.115–
41Agnes 2.5 Pro Beta–15.7%––
42GLM 5.3 Flashmax15.4%$0.042–
43Grok 4.5high15.4%$0.247–
44Muse Spark 1.1xhigh15.1%$0.376–
45Muse Spark Open (Xhigh)xhigh14.9%$0.271–
46Gemini 3.7 Flashhigh14.3%$0.115–
47DeepSeek V4.1 Flashmax14.3%$0.135–
48JT-4.1 Flash 236B A21B (Reasoning)–14.0%––
49Qwen 3.7 Maxmax13.4%$0.816–
50Gemini 3.5 Flashhigh13.1%$0.440–
51Beam 1.0–13.1%––
52DeepSeek V4 Pro 0424max12.9%$0.093–
53Opus 4.6max12.6%––
54Opus 4.7max12.0%––
55MiMo V2.6 Flash–12.0%$0.037–
56GPT-5.2xhigh11.6%––
57Muse Spark–11.3%––
58Qwen 3.8 Flash–11.1%$0.056–
59DeepSeek V4 Flashmax10.9%$0.111–
60Mistral Large 4–10.6%$0.354–
61Gemini 3.6 Flashhigh10.6%$0.135–
62Nemotron 3.5 Super VL–10.6%$0.054–
63Kimi K2.7 Code–10.0%$0.256–
64GPT-5.4 Minixhigh10.0%$0.560–
65Quasar 438B (Max, Based on GLM-5.2)max9.4%$0.358–
66GPT-5.4 Nanoxhigh9.3%$0.142–
67Gemini 3 Prohigh9.1%––
68Grok Build 0.1–9.1%––
69Qwen 3.7 Plus–9.1%$0.206–
70A.X-K2–8.9%––
71GPT-5.2 Codexxhigh8.7%––
72Nex-N2-Pro–8.6%––
73Gemini 3 Flash–8.6%––
74Inkling Small–8.3%$0.122–
75Kimi K2.6–8.0%$0.733–
76Grok 4.3high8.0%$0.127–
77DeepSeek V3.2–7.4%––
78DeepSeek V4 Flash 0420max7.1%$0.048–
79Motif 3–6.6%––
80Grok 4.20–6.6%––
81Solar Pro 4–6.3%$0.119–
82Motif 3 (Beta)–6.0%––
83Solar Open2 250B–5.7%––
84GPT-5.1 Codexhigh5.7%––
85GPT-5high5.7%––
86Qwen 3.8 27Bxhigh5.4%$0.322–
87Inklingxhigh5.4%––
88GPT-5 Codexhigh5.1%––
89Hy3–4.9%$0.037–
90GPT-5.1high4.9%––
91K2 Horizon 375B A23B–4.6%––
92Opus 4.5–4.6%––
93Apodex 1.1–4.6%$0.166–
94GLM 5.1–4.6%$0.362–
95MiMo V2 Flash–4.3%––
96Nemotron 3.5 Bolt 74B-A9B–4.3%––
97MiMo V2.5 Pro–4.0%$0.057–
98MiniMax M3–3.7%$0.096–
99MiMo V2.5–3.7%$0.012–
100Ring-2.6-1Txhigh3.7%$0.279–
101Qwen 3.6 Maxmax3.7%––
102Sonnet 4.6max3.1%$3.50–
103Kimi K2.5–3.1%––
104Nemotron 3 Ultramax3.1%$0.208–
105Nemotron 3 Super 120B A12B (Reasoning)–3.1%$0.055–
106Qwen 3.6 Plus–2.9%––
107Grok 4.1 Fast–2.9%––
108Grok 4 Fast–2.9%––
109GPT-5.5 Instant–2.6%––
110Ling 3.0 Flash Fin–2.6%––
111Kimi K2–2.6%––
112LongCat 2.0–2.6%$0.178–
113Muse Glimmer (High)high2.6%$0.036–
114Gemini 2.5 Pro–2.6%$0.267–
115Step 3.5 Flash–2.5%––
116K2 Horizon MoVA 36B A4B–2.3%––
117Step 3.7 Flashhigh2.3%––
118Step 3.5 Flash 2603–2.3%––
119GLM 5–2.0%––
120Ling 3.0 Flash–2.0%––
121Grok 4–2.0%––
122DeepSeek V3.1–2.0%––
123Qwen 3 Maxmax1.7%––
124Ling 3.0 Flash–1.7%––
125GLM 4.7–1.7%––
126Qwen 3.5 397B-A17B–1.7%$0.164–
127Solar Mini 4–1.4%$0.053–
128GPT-5 Minimedium1.4%––
129Gemini 2.5 Flashnone1.4%––
130Gemma 4 31B–1.4%$0.0000–
131ERNIE 5.0 Thinking Preview–1.4%––
132DeepSeek R1–1.4%––
133gpt-oss-20bhigh1.4%$0.0070–
134MiMo V2–1.1%––
135MiniMax M2.5–1.1%––
136Qwen 3.6 27B–1.1%$0.175–
137Sonnet 4.5–1.1%$0.452–
138o3–1.1%––
139Sonnet 4none1.1%––
140GLM 4.6–1.1%––
141Gemini 3.1 Flash-Litehigh1.1%$0.069–
142Apriel-v1.5-15B-Thinker–1.1%––
143gpt-oss-120bhigh1.1%$0.018–
144K-EXAONE (Reasoning)–1.1%––
145K-EXAONE 2.0 0803–0.9%––
146Qwen 3.5 35B-A3B–0.9%––
147MiniMax M2–0.9%––
148Qwen 3.5 122B-A10Bnone0.9%––
149Sonnet 3.7–0.9%––
150HyperNova 60B 2605 (High, Based on gpt-oss-120b)high0.9%––
151Trinity Large Thinking–0.9%$0.091–
152NVIDIA Nemotron 3 Nano 30B A3B (Reasoning)–0.9%$0.0090–
153Qwen 3.5 27B–0.9%––
154Mercury 2high0.8%$0.027–
155GLM 5 V–0.6%––
156MiniMax M2.7–0.6%––
157o4 minihigh0.6%––
158Grok 3 Minihigh0.6%––
159Qwen 3.5 9Bnone0.6%––
160Nemotron Cascade 2 30B A3B–0.6%––
161Ring-1T–0.6%––
162Ling-1Tnone0.6%––
163Gemma 4 E4B (Reasoning)–0.6%$0.0008–
164Qwen 3.5none0.6%––
165MiMo V2 Pro–0.3%––
166G9v3-39A5B–0.3%––
167MiniMax M2.1–0.3%––
168K2 Horizon 7B–0.3%––
169Qwen 3.6 35B-A3B–0.3%$0.071–
170Ling-2.6-1Tnone0.3%––
171Doubao Seed Code–0.3%––
172o1–0.3%––
173GLM 4.7 Flash–0.3%––
174Qwen 3–0.3%––
175Apriel-v1.6-15B-Thinker–0.3%––
176Nova 2.0 Lite (High)high0.3%––
177EXAONE 4.5 33B (Reasoning)–0.3%––
178Command A–0.3%$0.0000–
179Granite 4.2 30B–0.3%$0.016–
180MiniCPM5-2B–0.3%––
181Magistral Medium 1.2medium0.3%––
182Mistral Small 4–0.3%$0.036–
183Granite 4.2 8B–0.3%$0.0092–
184o3 minihigh0.3%––
185INTELLECT-3 (Based on GLM-4.5-Air)–0.3%––
186North Mini Code–0.3%$0.0000–
187Qwen 3 30B-A3B–0.3%$0.031–
188DiffusionGemma 26B A4B–0.3%––
189Magistral Medium 1medium0.3%––
190Tri-21B-Think–0.3%––
191Magistral Small 1.2–0.3%––
192Falcon-H1R-7B–0.3%––
193Llama 3.1 405b–0.3%––
194Gemma 4 E4B (Non-reasoning)none0.3%––
195Qwen 3 32B–0.3%––
196Sarvam 30B (High)high0.3%––
197Ring-flash-2.0–0.3%––
198JT-4.1 Flash 236B A21B (Non-reasoning)none0.0%––
199Opus 4.1–0.0%––
200Gemini 3.5 Flash-Litehigh0.0%$0.070–
201KAT Coder Pro V2none0.0%––
202GPT-5.1 Codex Minihigh0.0%––
203JT-35B-Flashnone0.0%––
204KAT-Coder-Pro V1none0.0%––
205Haiku 4.5–0.0%$0.165–
206Gemma 4 26B–0.0%––
207K2 Horizon 3.7B–0.0%––
208Mistral Medium 3.5high0.0%$0.483–
209Gemma 4 12B–0.0%––
210Nova 2.0 Pro Preview (Medium)medium0.0%––
211Grok Code Fast 1–0.0%––
212Nova 2.0 Omni (Medium)medium0.0%––
213Qwen 3.5 4B–0.0%––
214GPT-5 Nanohigh0.0%––
215Nemotron 3.5 Lightning–0.0%$0.0082–
216Nova 2.0 Pro Preview (Low)low0.0%––
217GLM 4.5–0.0%––
218Qwen 3 235B-A22B–0.0%––
219GPT-4.1none0.0%––
220Nova 2.0 Lite (Medium)medium0.0%––
221Mercury 2.5–0.0%$0.0088–
222JT-MINInone0.0%––
223Grok 3none0.0%––
224Seed-OSS-36B-Instruct–0.0%––
225Qwen 3 Codernone0.0%––
226Nova 2.0 Lite (Low)low0.0%––
227MiniMax M1–0.0%––
228K2 Think V2high0.0%––
229LongCat Flash Litenone0.0%––
230HyperCLOVA X SEED Think (32B)–0.0%––
231GLM 4.6 V–0.0%––
232K-EXAONE (Non-reasoning)none0.0%––
233Nova 2.0 Omni (Low)low0.0%––
234GLM 4.5 Air–0.0%––
235Ling 3.0 Tiny–0.0%$0.0000–
236Mi:dm K 2.5 Pro–0.0%––
237G9v3-3B–0.0%––
238Solar Open 100B (Reasoning)high0.0%––
239Gemini 2.5 Flash Lite–0.0%––
240Nemotron 3 Nano Omni 30B A3B Reasoning–0.0%––
241GPT-4.1 Mininone0.0%––
242Nova 2.0 Pro Preview (Non-reasoning)none0.0%––
243K2-V2 (High)high0.0%––
244DeepSeek V3none0.0%$0.0028–
245Ling 2.6 Flashnone0.0%––
246Tri-21B-think Preview–0.0%––
247Llama 4 Mavericknone0.0%$0.0024–
248Mistral Large 3none0.0%$0.0088–
249Motif-2-12.7B-Reasoning–0.0%––
250Mistral Medium 3.1none0.0%––
251Nova Premiernone0.0%––
252Granite 4.2 3B–0.0%$0.0031–
253Mistral Medium 3none0.0%––
254K2-V2 (Medium)medium0.0%––
255Llama Nemotron Super 49B v1.5 (Reasoning)–0.0%––
256Devstral Mediumnone0.0%––
257Gemini 2.0 Flashnone0.0%––
258Haiku 3.5none0.0%––
259Llama 3.3–0.0%––
260Qwen 3 4B–0.0%––
261MiniCPM5-1B (Reasoning)–0.0%––
262Sarvam 105B (High)high0.0%––
263Nova 2.0 Lite (Non-reasoning)none0.0%––
264Devstral Smallnone0.0%––
265MiniCPM5-1B (Non-reasoning)none0.0%––
266Devstral 2none0.0%––
267GPT-4onone0.0%––
268Nanbeige4.1-3B–0.0%––
269LFM2.5-2.6B–0.0%––
270Magistral Small 1–0.0%––
271EXAONE 4.0 32B (Reasoning)–0.0%––
272Nova 2.0 Omni (Non-reasoning)none0.0%––
273Llama 3.1 70b–0.0%––
274GPT-4.1 Nanonone0.0%––
275Solar Pro 3high0.0%$0.045–
276Ling-flash-2.0none0.0%––
277Gemma 4 E2B (Reasoning)–0.0%––
278Qwen 2.5 72Bnone0.0%––
279Step3 VL 10B–0.0%––
280Llama 3.3 70bnone0.0%––
281GLM 4.5 V–0.0%––
282Mistral Large 2none0.0%––
283Devstral Small 2none0.0%––
284Llama 3.1max0.0%––
285ERNIE 4.5 300B A47Bnone0.0%––
286Solar Pro 2 (Reasoning)high0.0%––
287NVIDIA Nemotron Nano 12B v2 VL (Reasoning)–0.0%––
288Granite 4.1 30Bnone0.0%––
289NVIDIA Nemotron Nano 9B V2 (Reasoning)–0.0%––
290Llama Nemotron Super 49B v1.5 (Non-reasoning)none0.0%––
291K2-V2 (Low)low0.0%––
292LFM2.5-8B-A1B–0.0%––
293Olmo 3.1 32B Think–0.0%––
294Mistral Small 3.1none0.0%$0.0006–
295Solar Pro 2 (Non-reasoning)none0.0%––
296Mistral Small 3.2none0.0%$0.0007–
297Nova Pronone0.0%––
298Qwen 3.5 2B–0.0%––
299Llama 3.1 8bnone0.0%––
300NVIDIA Nemotron 3 Nano 30B A3B (Non-reasoning)none0.0%––
301NVIDIA Nemotron Nano 9B V2 (Non-reasoning)none0.0%––
302Qwen 3 14Bnone0.0%––
303Nova Litenone0.0%––
304Granite 4.1 8Bnone0.0%––
305Olmo 3.1 32B Instructnone0.0%––
306Gemma 4 E2B (Non-reasoning)none0.0%––
307Olmo 3 32B Think–0.0%––
308Llama 4 Scoutnone0.0%$0.0018–
309Celeris-1none0.0%$0.0025–
310Phi-4 Mini Instructnone0.0%––
311EXAONE 4.0 32B (Non-reasoning)none0.0%––
312Qwen 3.5 0.8B–0.0%––
313Jamba 1.7 Largenone0.0%––
314Granite 4.0 H Smallnone0.0%––
315Ministral 3 14Bnone0.0%$0.0025–
316Qwen 3 8Bnone0.0%––
317LFM2 24B A2Bnone0.0%––
318Phi-4none0.0%––
319Nova Micronone0.0%––
320Granite 4.1 3Bnone0.0%––
321NVIDIA Nemotron Nano 12B v2 VL (Non-reasoning)none0.0%––
322MiniCPM-V 4.6 1.3Bnone0.0%––
323Jamba Reasoning 3B–0.0%––
324Reka Flash 3–0.0%––
325Olmo 3 7B Think–0.0%––
326Haiku 3none0.0%––
327Ling-mini-2.0none0.0%––
328Ministral 3 8Bnone0.0%$0.0025–
329Llama 3 70bnone0.0%––
330Llama 3.2 11bnone0.0%––
331Sarvam M (Reasoning, Based on Mistral Small 3.1)high0.0%––
332Exaone 4.0 1.2B (Reasoning)–0.0%––
333Olmo 3 7B Instructnone0.0%––
334Exaone 4.0 1.2B (Non-reasoning)none0.0%––
335LFM2.5-1.2B-Thinking–0.0%––
336Jamba 1.7 Mininone0.0%––
337LFM2 2.6Bnone0.0%––
338LFM2.5-1.2B-Instructnone0.0%––
339Granite 4.0 H 1Bnone0.0%––
340Qwen 3 1.7B–0.0%––
341Gemma 3 270Mnone0.0%––
342Apertus 70B Instructnone0.0%––
343Granite 4.0 Micronone0.0%––
344Mistralnone0.0%––
345Granite 4.0 1Bnone0.0%––
346Molmo2-8Bnone0.0%––
347LFM2 8B A1Bnone0.0%––
348Granite 3.3 8B (Non-reasoning)none0.0%––
349Gemma 3 27Bnone0.0%$0.0000–
350Ministral 3 3Bnone0.0%$0.0027–
351Qwen 3 0.6Bnone0.0%––
352LFM2 1.2Bnone0.0%––
353Gemma 3 4Bnone0.0%––
354LFM2.5-VL-1.6Bnone0.0%––
355Granite 4.0 350Mnone0.0%––
356Llama 3.2 1bnone0.0%––
357Tiny Aya Globalnone0.0%––
358Llama 3 8bnone0.0%––
359Apertus 8B Instructnone0.0%––
360Granite 4.0 H 350Mnone0.0%––
361Gemma 3n E4B Instructnone0.0%––
362Gemma 3n E2B Instructnone0.0%––
363Gemma 3 1Bnone0.0%––
364NVIDIA Nemotron 3 Nano 4B–0.0%––
365Gemma 3 12Bnone0.0%$0.0000–
366K2 Horizon 0.9B–0.0%––
367Mi:dm K 2.5 Pro Preview–0.0%––
368Cogito v2.1 (Reasoning)–0.0%––