Knowledge & reasoning

CritPt leaderboard

Name: benchgap: LLM benchmark scores, measured and estimated
Creator: benchgap

As of 2026-10-07, the highest measured score on CritPt is 32.3% by GPT-5.6 Sol. 52 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	GPT-5.6 Sol	32.3%	measured
2	Claude Opus 5.5	31.7%	measured
3	GPT-6.1 Sol	31.7%	measured
4	GPT-6 Astra	31.7%	measured
5	Claude Sonnet 5.5	31.4%	measured
6	GPT-6 Sol	30.9%	measured
7	GPT-5.5 Pro	30.6%	measured
8	Sakana Fugu	30.3%	estimated ± 4.1 pp, high confidence
9	Sakana Fugu-Ultra	30.3%	estimated ± 4.1 pp, high confidence
10	GPT-5.4 Pro	30.0%	measured
11	GPT-5.6 Terra	30.0%	measured
12	Claude Fable 5.1	29.7%	measured
13	Claude Opus 5	29.1%	measured
14	Claude Fable 5	28.6%	measured
15	Gemini 4 Argon	27.1%	measured
16	GPT-5.5	27.1%	measured
17	MiMo-V2.6-Pro	26.6%	measured
18	Gemini 3 Pro Deep Think	25.7%	measured
19	Muse Spark 1.3	24.9%	measured
20	Claude Mythos 5	24.6%	estimated ± 4.1 pp, high confidence
21	GPT-5.4	23.4%	measured
22	Kimi K3	23.4%	measured
23	dots3-note Preview	23.0%	estimated ± 5.7 pp, medium confidence
24	Pareto 26.9	21.6%	estimated ± 6.2 pp, medium confidence
25	Claude Opus 4.8	20.9%	measured
26	GLM-5.2	20.9%	measured
27	Step 5 Preview	20.9%	measured
28	GPT-5.6 Luna	20.6%	measured
29	GPT-6 Luna	19.4%	measured
30	GLM-5.3	19.1%	measured
31	Claude Haiku 5.5	18.9%	measured
32	Ornith-1.5-397B	18.8%	estimated ± 4.1 pp, high confidence
33	Gemini 3.8 Flash	18.3%	measured
34	DeepSeek V4 Pro 0813	18.0%	measured
35	Ling 3.1 Flash	18.0%	measured
36	Qwen3.8 Max	18.0%	estimated ± 4.1 pp, high confidence
37	Gemini 3.1 Pro	17.7%	measured
38	Grok 4.7	17.7%	measured
39	Muse Spark 1.2	17.7%	measured
40	Qwen3.8 Max Preview	17.7%	measured
41	Pareto 26.10 Preview	17.5%	estimated ± 4.9 pp, high confidence
42	Grok 4.6	17.1%	measured
43	Claude Sonnet 5	16.9%	measured
44	GPT-5.3 Codex	16.9%	measured
45	Hy4 preview	16.9%	measured
46	DeepSeek V4 Flash 0731	16.6%	measured
47	GLM-5.3-Flash	15.4%	measured
48	Grok 4.5	15.4%	measured
49	Muse Spark 1.1	15.1%	measured
50	DeepSeek V4.1 Flash	14.3%	measured
51	Gemini 3.7 Flash	14.3%	measured
52	Qwen3.7 Max	13.4%	measured
53	Gemini 3.5 Flash	13.1%	measured
54	Claude Opus 4.6 (Adaptive)	12.6%	measured
55	Beam	12.2%	estimated ± 4.9 pp, high confidence
56	Claude Opus 4.7 (Adaptive)	12.0%	measured
57	MiMo-V2.6-Flash	12.0%	measured
58	Agents-A1	12.0%	estimated ± 6.5 pp, medium confidence
59	Qwen3.8-Omni-Flash	11.6%	estimated ± 4.1 pp, high confidence
60	GPT-5.2	11.6%	measured
61	Muse Spark	11.3%	measured
62	Qwen3.8-Flash-Next	11.1%	measured
63	Gemini 3.6 Flash	10.6%	measured
64	Mistral Large 4	10.6%	measured
65	GPT-5.4 mini	10.0%	measured
66	Kimi K2.7 Code	10.0%	measured
67	Claude Sonnet 4.5 Thinking	9.8%	estimated ± 5.7 pp, medium confidence
68	Quasar 438B	9.4%	measured
69	GPT-5.4 nano	9.3%	measured
70	Gemini 3 Pro	9.1%	measured
71	Qwen3.7 Plus	9.1%	measured
72	A.X K2	8.9%	measured
73	GPT-5.2-Codex	8.7%	measured
74	Inkling-Small	8.3%	measured
75	Interfaze Beta	8.2%	estimated ± 4.1 pp, high confidence
76	Grok 4.3	8.0%	measured
77	Kimi K2.6	8.0%	measured
78	Grok 4.20	7.8%	estimated ± 4.9 pp, high confidence
79	Ornith-1.5-35B-A3B	6.5%	estimated ± 4.1 pp, high confidence
80	GPT-5.1-Codex	5.7%	measured
81	GPT-5.1-Codex-Max	5.7%	measured
82	GPT-5 (high)	5.7%	measured
83	Inkling	5.4%	measured
84	Qwen3.8-27B	5.4%	measured
85	Solar Pro 4	5.4%	measured
86	Claude Opus 4.7	5.1%	measured
87	GPT-5.1	4.9%	measured
88	Hy3	4.9%	measured
89	Hy3 Preview	4.9%	measured
90	Apodex 1.1	4.6%	measured
91	Apodex 1.1 Mini	4.6%	measured
92	Claude Opus 4.5 Thinking	4.6%	measured
93	GLM-5.1	4.6%	measured
94	Qwen3.5 Flash	4.0%	estimated ± 6.6 pp, medium confidence
95	MiMo-V2.5-Pro	4.0%	measured
96	Solar Open 2	3.8%	estimated ± 4.9 pp, high confidence
97	MiniMax M3	3.7%	measured
98	Qwen 3.6 Max (preview)	3.7%	measured
99	MiMo-V2.5	3.6%	estimated ± 6.6 pp, medium confidence
100	Gemini 3.1 Flash-Lite	3.4%	estimated ± 6.6 pp, medium confidence
101	Kimi K2.5	3.1%	measured
102	Kimi K2.5 (Reasoning)	3.1%	measured
103	Nemotron 3 Super 100B	3.1%	measured
104	Nemotron 3 Ultra	3.1%	measured
105	Grok 4.1 Fast (Reasoning)	2.9%	measured
106	Grok 4 Fast (Reasoning)	2.9%	measured
107	Qwen3.6 Plus	2.9%	measured
108	Claude Opus 4.6	2.8%	measured
109	Claude Haiku 4.5	2.6%	estimated ± 6.6 pp, medium confidence
110	GLM-4.5	2.6%	estimated ± 6.6 pp, medium confidence
111	Gemini 2.5 Pro	2.6%	measured
112	Muse Glimmer 30B	2.6%	measured
113	Ornith-1.5-9B	2.6%	estimated ± 4.1 pp, high confidence
114	Laguna XS.2	2.6%	estimated ± 6.6 pp, medium confidence
115	Laguna M.1	2.6%	estimated ± 6.6 pp, low confidence
116	Step 3.7 Flash	2.3%	measured
117	Ternary Bonsai 2 27B	2.1%	estimated ± 4.1 pp, high confidence
118	DeepSeek V3.1 (Reasoning)	2.0%	measured
119	GLM-5	2.0%	measured
120	Grok 4	2.0%	measured
121	Ling 3.0 Flash VL	2.0%	measured
122	Claude 4.1 Opus	1.7%	estimated ± 3.7 pp, high confidence
123	GLM-4.7	1.7%	measured
124	Ling 3.0 Flash	1.7%	measured
125	Ling 3.0 Flash FP8	1.7%	measured
126	o3-pro	1.6%	estimated ± 2.9 pp, high confidence
127	DeepSeek-R1	1.4%	measured
128	Gemini 2.5 Flash	1.4%	measured
129	Gemini 3 Flash	1.4%	measured
130	Gemma 4 31B	1.4%	measured
131	GPT-OSS 20B	1.4%	measured
132	MAI-Thinking-1	1.4%	estimated ± 4.1 pp, high confidence
133	Claude Sonnet 4.5	1.1%	estimated ± 4.1 pp, high confidence
134	Claude 4 Sonnet	1.1%	measured
135	GPT-OSS 120B	1.1%	measured
136	K-Exaone	1.1%	measured
137	MiMo-V2-Omni	1.1%	measured
138	o3	1.1%	measured
139	Qwen3.6-27B	1.1%	measured
140	Claude Sonnet 4.6	0.9%	measured
141	DeepSeek V3.2	0.9%	measured
142	K-EXAONE 2.0	0.9%	measured
143	Nemotron 3 Nano 30B	0.9%	measured
144	Qwen3.5-27B	0.9%	measured
145	Qwen3.5-35B-A3B	0.9%	measured
146	Qwen3.5 397B	0.9%	measured
147	Qwen3.5 397B (Reasoning)	0.9%	measured
148	Trinity-Large-Preview	0.9%	measured
149	Trinity-Large-Thinking	0.9%	measured
150	Qwen3 235B 2507	0.6%	estimated ± 4.1 pp, high confidence
151	Gemma 4 E4B	0.6%	measured
152	GLM-5V-Turbo	0.6%	measured
153	MiniMax M2.7	0.6%	measured
154	Qwen3.5-122B-A10B	0.6%	measured
155	o1-preview	0.6%	estimated ± 2.9 pp, high confidence
156	ZAYA1-8B	0.6%	estimated ± 4.1 pp, high confidence
157	Claude 3.5 Sonnet	0.6%	estimated ± 4.1 pp, high confidence
158	LFM2.5-230M	0.6%	estimated ± 4.1 pp, medium confidence
159	LFM2.5-VL-450M	0.6%	estimated ± 4.1 pp, medium confidence
160	Mellum2-12B-A2.5B-Instruct	0.6%	estimated ± 4.1 pp, medium confidence
161	Mellum2-12B-A2.5B-Thinking	0.6%	estimated ± 4.1 pp, high confidence
162	Soofi S 30B-A3B	0.6%	estimated ± 4.1 pp, high confidence
163	ZAYA1-74B-Preview	0.6%	estimated ± 4.1 pp, high confidence
164	o1-pro	0.4%	estimated ± 3.7 pp, high confidence
165	o3-mini	0.3%	estimated ± 2.3 pp, high confidence
166	Claude Opus 4.5	0.3%	measured
167	Command A+	0.3%	measured
168	GLM-5-Turbo	0.3%	measured
169	Granite 4.2 30B	0.3%	measured
170	Granite 4.2 8B	0.3%	measured
171	MiMo-V2-Pro	0.3%	measured
172	MiniCPM5-2B	0.3%	measured
173	Mistral Small 4	0.3%	measured
174	Mistral Small 4 (Reasoning)	0.3%	measured
175	North Mini Code	0.3%	measured
176	o1	0.3%	measured
177	Qwen3.6-35B-A3B	0.3%	measured
178	Sarvam 30B	0.3%	measured
179	Phi-4 Multimodal Instruct	0.3%	estimated ± 2.3 pp, high confidence
180	DeepSeek R1 Distill Qwen 32B	0.3%	estimated ± 2.3 pp, high confidence
181	Gemini 1.5 Pro	0.3%	estimated ± 2.3 pp, high confidence
182	Gemini 1.0 Pro	0.3%	estimated ± 2.3 pp, high confidence
183	GPT-4o mini	0.3%	estimated ± 2.3 pp, high confidence
184	Qwen2.5 Coder 32B Instruct	0.3%	estimated ± 2.3 pp, high confidence
185	GPT-4 Turbo	0.3%	estimated ± 2.3 pp, high confidence
186	Claude 3 Opus	0.3%	estimated ± 2.3 pp, high confidence
187	LongCat-Flash-Lite-Sparse	0.1%	estimated ± 1.1 pp, low confidence
188	Celeris-1	0.0%	measured
189	Claude 3 Haiku	0.0%	measured
190	Claude 4.1 Opus Thinking	0.0%	measured
191	DeepSeek V3	0.0%	measured
192	DeepSeek V3 0324	0.0%	measured
193	DeepSeek V3.1	0.0%	measured
194	Exaone 4.0 1.2B	0.0%	measured
195	Exaone 4.0 32B	0.0%	measured
196	Gemini 3.5 Flash-Lite	0.0%	measured
197	Gemma 3 27B	0.0%	measured
198	Gemma 4 12B	0.0%	measured
199	Gemma 4 26B A4B	0.0%	measured
200	Gemma 4 E2B	0.0%	measured
201	GLM-4.5-Air	0.0%	measured
202	GLM-4.6	0.0%	measured
203	GPT-4.1	0.0%	measured
204	GPT-4.1 mini	0.0%	measured
205	GPT-4.1 nano	0.0%	measured
206	GPT-4o	0.0%	measured
207	GPT-5 (medium)	0.0%	measured
208	Granite-4.0-350M	0.0%	measured
209	Granite-4.0-H-1B	0.0%	measured
210	Granite-4.0-H-350M	0.0%	measured
211	Granite 4.2 3B	0.0%	measured
212	Grok 4.1 Fast	0.0%	measured
213	Grok Code Fast 1	0.0%	measured
214	Kimi K2	0.0%	measured
215	LFM2.5-2.6B	0.0%	measured
216	LFM2.5-8B-A1B	0.0%	measured
217	LFM2.5-VL-1.6B-Extract	0.0%	measured
218	Ling 2.6 Flash	0.0%	measured
219	Ling 3.0 Tiny	0.0%	measured
220	LLaDA2.2-mini	0.0%	estimated ± 4.9 pp, medium confidence
221	Llama 3.1 405B	0.0%	measured
222	Llama 4 Maverick	0.0%	measured
223	Llama 4 Scout	0.0%	measured
224	Mercury 2.5	0.0%	measured
225	MiMo-V2-Flash	0.0%	measured
226	MiniCPM5-1B	0.0%	estimated ± 4.9 pp, medium confidence
227	Mistral Large 2	0.0%	measured
228	Mistral Large 3	0.0%	measured
229	Mistral Medium 3	0.0%	measured
230	Mistral Medium 3.5 128B	0.0%	measured
231	Nemotron 3.5 Lightning 30B A3B NVFP4	0.0%	measured
232	Nemotron 3 Nano Omni 30B A3B	0.0%	measured
233	Nemotron Ultra 253B	0.0%	measured
234	Nova Pro	0.0%	measured
235	Phi-4	0.0%	measured
236	Qwen3 Max	0.0%	measured
237	Qwen3-Omni-30B-A3B-Instruct	0.0%	measured
238	Qwen3-Omni-30B-A3B-Thinking	0.0%	measured
239	Sarvam 105B	0.0%	measured
240	Solar Pro 2	0.0%	measured
241	Solar Pro 3	0.0%	measured
242	Ultravox v0.6 Llama 3.3 70B	0.0%	measured

CritPt leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General