benchgap
Knowledge & reasoning

CritPt leaderboard

As of 2026-11-21, the highest measured score on CritPt is 35.4% by Claude Fable 5.2. 104 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1Claude Fable 5.235.4%measured
2GPT-6.2 Astra35.3%estimated ± 4.0 pp, medium confidence
3Muse Wonder Full32.9%measured
4GPT-5.6 Sol32.3%measured
5GPT-6.2 Sol32.1%estimated ± 4.0 pp, high confidence
6Claude Opus 5.531.7%measured
7GPT-6.1 Sol31.7%measured
8GPT-6 Astra31.7%measured
9Claude Sonnet 5.531.4%measured
10Gemini 4 Neon31.0%estimated ± 4.7 pp, high confidence
11GPT-6 Sol30.9%measured
12Gemini 4 Carbon30.7%estimated ± 4.7 pp, high confidence
13GPT-5.5 Pro30.6%measured
14Sakana Fugu30.3%estimated ± 4.0 pp, high confidence
15Sakana Fugu-Ultra30.3%estimated ± 4.0 pp, high confidence
16GPT-5.4 Pro30.0%measured
17GPT-5.6 Terra30.0%measured
18Claude Fable 5.129.7%measured
19Claude Opus 529.1%measured
20Gemini 4 Hydrogen28.7%estimated ± 4.7 pp, high confidence
21Claude Fable 528.6%measured
22Muse Wonder28.3%measured
23GPT-6.1 Terra28.0%measured
24Gemini 4 Argon27.1%measured
25GPT-5.527.1%measured
26MiMo-V2.6-Pro26.6%measured
27Muse Spark 1.324.9%measured
28Kimi K3.1 Pro24.6%measured
29Claude Mythos 524.5%estimated ± 4.0 pp, high confidence
30dots3-note Preview23.5%estimated ± 5.1 pp, medium confidence
31GPT-5.423.4%measured
32Kimi K323.4%measured
33MiniMax M3.1 Flash22.3%measured
34Pareto 26.921.9%estimated ± 5.5 pp, medium confidence
35Qwen4 Max21.6%estimated ± 4.7 pp, high confidence
36Claude Opus 4.820.9%measured
37GLM-5.220.9%measured
38Step 5 Preview20.9%measured
39GPT-5.6 Luna20.6%measured
40ERNIE 5.5 Flash Preview20.3%estimated ± 5.5 pp, medium confidence
41Hy420.2%estimated ± 6.1 pp, medium confidence
42Sage-A1.520.0%measured
43GPT-6 Luna19.4%measured
44Inkling 2 Small19.1%estimated ± 3.4 pp, high confidence
45GLM-5.319.1%measured
46Claude Haiku 5.518.9%measured
47Ornith-1.5-397B18.8%estimated ± 4.0 pp, high confidence
48Hy4 flash18.4%estimated ± 6.1 pp, medium confidence
49Gemini 3.8 Flash18.3%measured
50Pareto 26.10 Preview18.1%estimated ± 4.6 pp, high confidence
51DeepSeek V4 Pro18.0%measured
52Ling 3.1 Flash18.0%measured
53Qwen3.8 Max17.9%estimated ± 4.0 pp, high confidence
54Gemini 3.1 Pro17.7%measured
55Grok 4.717.7%measured
56Muse Spark 1.217.7%measured
57Qwen3.8 Max Preview17.7%measured
58Beam 1.0 501B-A23B FP817.4%measured
59Grok 4.617.1%measured
60Claude Sonnet 516.9%measured
61GPT-5.3 Codex16.9%measured
62Hy4 preview16.9%measured
63DeepSeek V4 Flash 073116.6%measured
64Beam 1.0 501B-A23B16.3%measured
65Grok 4.515.4%measured
66Beam 1.0 501B-A23B NVFP415.4%measured
67Muse Ember15.3%estimated ± 4.7 pp, high confidence
68Muse Spark 1.115.1%measured
69Muse Spark Open14.9%measured
70Agents-A114.8%estimated ± 6.1 pp, medium confidence
71DeepSeek V4.1 Flash 101714.3%measured
72Gemini 3.7 Flash14.3%measured
73Qwen3.7 Max13.4%measured
74Gemini 3.5 Flash13.1%measured
75Particle-112.7%estimated ± 4.6 pp, high confidence
76Claude Opus 4.6 (Adaptive)12.6%measured
77Interfaze Beta12.2%estimated ± 2.9 pp, low confidence
78Claude Opus 4.712.0%measured
79MiMo-V2.6-Flash12.0%measured
80Qwen3.8-Omni-Flash11.7%estimated ± 4.0 pp, high confidence
81GPT-5.211.6%measured
82Muse Spark11.3%measured
83Qwen3.8-Flash-Next11.1%measured
84Nemotron 3.5 Super VL 120B A12B10.9%measured
85Gemini 3.6 Flash10.6%measured
86Nemotron 3.5 Super VL 120B A12B NVFP410.6%measured
87GPT-5.4 mini10.0%measured
88Kimi K2.7 Code10.0%measured
89Mistral Large 410.0%measured
90Mercury 39.9%estimated ± 4.6 pp, high confidence
91Quasar 438B9.4%measured
92GPT-5.4 nano9.3%measured
93Gemini 3 Pro9.1%measured
94Qwen3.7 Plus9.1%measured
95Nemotron 3.5 Bolt 74B-A9B BF169.0%estimated ± 4.6 pp, high confidence
96A.X K28.9%measured
97GPT-5.2-Codex8.7%measured
98Claude Sonnet 4.5 Thinking8.7%estimated ± 5.1 pp, medium confidence
99Inkling-Small8.3%measured
100Grok 4.208.3%estimated ± 4.6 pp, high confidence
101Grok 4.38.0%measured
102Kimi K2.68.0%measured
103OrcaCyber-Nano-1.07.8%estimated ± 4.6 pp, high confidence
104Ornith-1.5-35B-A3B6.7%estimated ± 4.0 pp, high confidence
105Claude 4 Opus Thinking6.4%estimated ± 2.7 pp, high confidence
106Solar Pro 46.3%measured
107GLM-5.3-Flash5.9%estimated ± 5.3 pp, medium confidence
108Claude 4 Opus5.8%estimated ± 2.9 pp, medium confidence
109GPT-55.7%measured
110GPT-5.1-Codex5.7%measured
111GPT-5.1-Codex-Max5.7%measured
112Inkling5.4%measured
113Qwen3.8-27B5.4%measured
114GPT-5.14.9%measured
115Hy34.9%measured
116Hy3 Preview4.9%measured
117Apodex 1.14.6%measured
118Apodex 1.1 Mini4.6%measured
119Claude Opus 4.5 Thinking4.6%measured
120GLM-5.14.6%measured
121Nemotron 3.5 Bolt 74B-A9B4.3%measured
122MiMo-V2.5-Pro4.0%measured
123Solar Open 23.7%estimated ± 4.1 pp, medium confidence
124MiniMax M33.7%measured
125Qwen 3.6 Max (preview)3.7%measured
126Claude 4.1 Opus3.4%estimated ± 4.7 pp, high confidence
127Kimi K2.53.1%measured
128Nemotron 3 Ultra3.1%measured
129Grok 4.1 Fast (Reasoning)2.9%measured
130Grok 4 Fast (Reasoning)2.9%measured
131Qwen3.6 Plus2.9%measured
132Claude Opus 4.62.8%measured
133Inkling 2 Mini2.8%estimated ± 4.0 pp, high confidence
134Ornith-1.5-9B2.6%estimated ± 4.0 pp, high confidence
135Gemini 2.5 Pro2.6%measured
136Muse Glimmer 30B2.6%measured
137Qwen3.5 Flash2.5%estimated ± 5.3 pp, medium confidence
138Step 3.7 Flash2.3%measured
139Ternary Bonsai 2 27B2.1%estimated ± 4.0 pp, high confidence
140GLM-52.0%measured
141Grok 42.0%measured
142Ling 3.0 Flash VL2.0%measured
143MiMo-V2.52.0%estimated ± 5.3 pp, medium confidence
144o3-pro1.8%estimated ± 3.8 pp, high confidence
145Gemini 3.1 Flash-Lite1.8%estimated ± 5.3 pp, medium confidence
146GLM-4.71.7%measured
147Ling 3.0 Flash1.7%measured
148Qwen3.5 397B1.7%measured
149DeepSeek-R11.4%measured
150Gemini 2.5 Flash1.4%measured
151Gemini 3 Flash1.4%measured
152Gemma 4 31B1.4%measured
153GPT-OSS 20B1.4%measured
154MAI-Thinking-11.3%estimated ± 4.0 pp, high confidence
155Ling 3.0 Flash FP81.2%estimated ± 4.0 pp, high confidence
156Ripple-65B Preview1.2%estimated ± 4.1 pp, medium confidence
157Claude 4 Sonnet1.1%measured
158GPT-OSS 120B1.1%measured
159K-Exaone1.1%measured
160MiMo-V2-Omni1.1%measured
161o31.1%measured
162Qwen3.6-27B1.1%measured
163Claude Sonnet 4.51.0%estimated ± 4.0 pp, high confidence
164Claude 3.7 Sonnet Thinking0.9%measured
165Claude Sonnet 4.60.9%measured
166DeepSeek V3.20.9%measured
167K-EXAONE 2.00.9%measured
168Qwen3.5-27B0.9%measured
169Qwen3.5-35B-A3B0.9%measured
170Trinity-Large-Preview0.9%measured
171Trinity-Large-Thinking0.9%measured
172Claude Haiku 4.50.9%estimated ± 5.3 pp, medium confidence
173GLM-4.50.9%estimated ± 5.3 pp, medium confidence
174Laguna XS.20.8%estimated ± 5.3 pp, medium confidence
175Laguna M.10.8%estimated ± 5.3 pp, low confidence
176DeepSeek-R1-05280.6%estimated ± 4.0 pp, high confidence
177Gemma 4 E4B0.6%measured
178GLM-5V-Turbo0.6%measured
179MiniMax M2.70.6%measured
180o4-mini0.6%measured
181Qwen3.5-122B-A10B0.6%measured
182Grok 3 Mini0.5%estimated ± 4.0 pp, high confidence
183o1-preview0.5%estimated ± 3.8 pp, high confidence
184Nemotron 3 Super 100B0.5%estimated ± 4.0 pp, high confidence
185o1-pro0.5%estimated ± 4.0 pp, high confidence
186DeepSeek-V2.50.4%estimated ± 3.8 pp, high confidence
187Qwen3 235B 25070.4%estimated ± 4.0 pp, high confidence
188Grok 3 [Beta]0.4%estimated ± 4.0 pp, high confidence
189GLM-4.7-Flash0.4%estimated ± 4.0 pp, high confidence
190LFM2.5-24B-A2B0.3%estimated ± 4.0 pp, high confidence
191ZAYA1-8B0.3%estimated ± 4.0 pp, high confidence
192Spark-X2.6-4B0.3%estimated ± 4.0 pp, high confidence
193GPT-4.50.3%estimated ± 4.0 pp, high confidence
194Phi-4-reasoning-plus0.3%estimated ± 4.0 pp, high confidence
195DeepSeek R1 Distill Llama 70B0.3%estimated ± 4.0 pp, high confidence
196LFM3-1.7B0.3%estimated ± 4.0 pp, high confidence
197DeepSeek-R1-0528-Qwen3-8B0.3%estimated ± 4.0 pp, high confidence
198Qwen2.5-Max0.3%estimated ± 4.0 pp, high confidence
199Claude 3.5 Sonnet0.3%estimated ± 4.0 pp, high confidence
200DeepSeek R1 Distill Qwen 14B0.3%estimated ± 4.0 pp, high confidence
201Mellum2-12B-A2.5B-Thinking0.3%estimated ± 4.0 pp, high confidence
202ZAYA1-74B-Preview0.3%estimated ± 4.0 pp, high confidence
203Grok 20.3%estimated ± 4.0 pp, high confidence
204MiniMax-Text-010.3%estimated ± 4.0 pp, high confidence
205Grok 2 mini0.3%estimated ± 4.0 pp, high confidence
206Gemini 2.0 Flash-Lite0.3%estimated ± 4.0 pp, high confidence
207Qwen2.5-VL-32B0.3%estimated ± 4.0 pp, high confidence
208Soofi S 30B-A3B0.3%estimated ± 4.0 pp, high confidence
209Mellum2-12B-A2.5B-Instruct0.3%estimated ± 4.0 pp, high confidence
210LFM2.5-350M0.3%estimated ± 4.0 pp, high confidence
211Phi-3.5-mini0.3%estimated ± 4.0 pp, high confidence
212Granite-4.0-1B0.3%estimated ± 4.0 pp, high confidence
213LFM2.5-VL-450M0.3%estimated ± 4.0 pp, high confidence
214LFM2.5-230M0.3%estimated ± 4.0 pp, high confidence
215o3-mini0.3%estimated ± 0.3 pp, high confidence
216Claude Opus 4.50.3%measured
217Command A+0.3%measured
218GLM-5-Turbo0.3%measured
219Granite 4.2 30B0.3%measured
220Granite 4.2 8B0.3%measured
221Magistral Medium0.3%measured
222MiMo-V2-Pro0.3%measured
223MiniCPM5-2B0.3%measured
224Mistral Small 40.3%measured
225North Mini Code0.3%measured
226o10.3%measured
227Qwen3.6-35B-A3B0.3%measured
228Sarvam 30B0.3%measured
229Phi-4 Multimodal Instruct0.2%estimated ± 3.4 pp, high confidence
230DeepSeek R1 Distill Qwen 32B0.2%estimated ± 3.4 pp, high confidence
231Gemini 1.5 Pro0.2%estimated ± 3.4 pp, high confidence
232Gemini 1.0 Pro0.2%estimated ± 3.4 pp, high confidence
233Qwen2.5 Coder 32B Instruct0.2%estimated ± 3.4 pp, high confidence
234Gemini 1.5 Flash0.1%estimated ± 3.4 pp, high confidence
235GPT-4 Turbo0.1%estimated ± 3.4 pp, high confidence
236Claude 3 Opus0.1%estimated ± 3.4 pp, high confidence
237GPT-4o mini0.0%estimated ± 0.3 pp, high confidence
238Celeris-10.0%measured
239Claude 3.5 Haiku0.0%measured
240Claude 3.7 Sonnet0.0%measured
241Claude 3 Haiku0.0%measured
242DeepSeek V30.0%measured
243DeepSeek V3 03240.0%measured
244DeepSeek V3.10.0%measured
245Devstral Small0.0%measured
246Exaone 4.0 1.2B0.0%measured
247Exaone 4.0 32B0.0%measured
248Gemini 2.0 Flash0.0%measured
249Gemini 2.5 Flash-Lite0.0%measured
250Gemini 3.5 Flash-Lite0.0%measured
251Gemma 3 27B0.0%measured
252Gemma 4 12B0.0%measured
253Gemma 4 26B A4B0.0%measured
254Gemma 4 E2B0.0%measured
255GLM-4.5-Air0.0%measured
256GLM-4.60.0%measured
257GPT-4.10.0%measured
258GPT-4.1 mini0.0%measured
259GPT-4.1 nano0.0%measured
260GPT-4o0.0%measured
261Granite-4.0-350M0.0%measured
262Granite-4.0-H-1B0.0%measured
263Granite-4.0-H-350M0.0%measured
264Granite 4.2 3B0.0%measured
265Grok 4.1 Fast0.0%measured
266Grok Code Fast 10.0%measured
267Kimi K20.0%measured
268LFM2.5-2.6B0.0%measured
269LFM2.5-8B-A1B0.0%measured
270LFM2.5-VL-1.6B-Extract0.0%measured
271Ling 2.6 Flash0.0%measured
272Ling 3.0 Tiny0.0%measured
273LLaDA2.2-mini0.0%estimated ± 4.6 pp, medium confidence
274Llama 3.1 405B0.0%measured
275Llama 3.1 70B0.0%measured
276Llama 3.1 8B0.0%estimated ± 0.3 pp, high confidence
277Llama 3.3 70B0.0%measured
278Llama 4 Maverick0.0%measured
279Llama 4 Scout0.0%measured
280LongCat-Flash-Lite-Sparse0.0%estimated ± 4.1 pp, medium confidence
281Magistral Small0.0%measured
282Mellum2.1-12B-A2.5B-Thinking0.0%estimated ± 4.6 pp, high confidence
283Mercury 2.50.0%measured
284MiMo-V2-Flash0.0%measured
285MiniCPM5-1B0.0%estimated ± 4.1 pp, medium confidence
286Mistral Large 20.0%measured
287Mistral Large 30.0%measured
288Mistral Medium 30.0%measured
289Mistral Medium 3.5 128B0.0%measured
290Mistral Small 3.10.0%measured
291Nemotron 3.5 Lightning 30B A3B NVFP40.0%measured
292Nemotron 3 Nano 30B0.0%measured
293Nemotron 3 Nano Omni 30B A3B0.0%measured
294Nemotron Ultra 253B0.0%measured
295Nova Pro0.0%measured
296o1-mini0.0%estimated ± 0.3 pp, high confidence
297Phi-40.0%measured
298Phi-4-reasoning0.0%measured
299Qwen3 235B-A22B0.0%estimated ± 0.8 pp, low confidence
300Qwen3 235B-A22B (Reasoning)0.0%estimated ± 0.8 pp, low confidence
301Qwen3 32B0.0%estimated ± 2.9 pp, low confidence
302Qwen3 32B (Reasoning)0.0%estimated ± 0.8 pp, low confidence
303Qwen3.5-4B0.0%measured
304Qwen3 Max0.0%measured
305Qwen3-Omni-30B-A3B-Instruct0.0%measured
306Qwen3-Omni-30B-A3B-Thinking0.0%measured
307QwQ-32B0.0%estimated ± 0.8 pp, low confidence
308Sarvam 105B0.0%measured
309Solar Pro 20.0%measured
310Solar Pro 30.0%measured
311Ultravox v0.6 Llama 3.3 70B0.0%measured

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General