Knowledge & reasoning

SuperGPQA leaderboard

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

As of 2026-10-07, the highest measured score on SuperGPQA is 95.0% by Claude Opus 4.6. 214 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	Claude Opus 4.6	95.0%	measured
2	Claude Sonnet 4.6	95.0%	measured
3	Claude Haiku 5.5	91.4%	estimated ± 10.4 pp, low confidence
4	Claude Fable 5.1	90.9%	estimated ± 7.5 pp, low confidence
5	Claude Opus 5.5	90.7%	estimated ± 7.5 pp, low confidence
6	Claude Fable 5	90.5%	estimated ± 7.5 pp, low confidence
7	GLM-5.3-Flash	90.3%	estimated ± 10.4 pp, low confidence
8	GPT-6 Astra	89.7%	estimated ± 7.5 pp, low confidence
9	GPT-6.1 Sol	89.6%	estimated ± 7.5 pp, low confidence
10	Claude Opus 5	89.3%	estimated ± 7.5 pp, low confidence
11	GPT-5.6 Sol	88.8%	estimated ± 7.5 pp, low confidence
12	GPT-5.5	88.4%	estimated ± 7.5 pp, low confidence
13	Gemini 3 Pro	87.7%	estimated ± 7.5 pp, low confidence
14	Claude Mythos 5	87.6%	estimated ± 9.1 pp, low confidence
15	Gemini 3.7 Flash	87.6%	estimated ± 7.5 pp, low confidence
16	Gemini 3.1 Pro	87.4%	estimated ± 7.5 pp, low confidence
17	Gemini 3.8 Flash	87.3%	estimated ± 7.5 pp, low confidence
18	GPT-6 Sol	87.3%	estimated ± 7.5 pp, low confidence
19	Claude Sonnet 5.5	87.1%	estimated ± 7.5 pp, low confidence
20	GPT-5.3 Codex	86.8%	estimated ± 7.5 pp, low confidence
21	Muse Spark 1.1	86.5%	estimated ± 7.5 pp, low confidence
22	Gemini 3.5 Flash	86.4%	estimated ± 7.5 pp, low confidence
23	Grok 4.5	86.3%	estimated ± 7.5 pp, low confidence
24	GPT-5.4	86.0%	estimated ± 7.5 pp, low confidence
25	GPT-5.4 Pro	85.9%	estimated ± 9.1 pp, low confidence
26	Gemini 3.6 Flash	85.7%	estimated ± 7.5 pp, low confidence
27	Gemini 4 Argon	85.7%	estimated ± 7.5 pp, low confidence
28	Muse Spark	85.5%	estimated ± 7.5 pp, low confidence
29	GPT-5.5 Pro	85.4%	estimated ± 9.1 pp, low confidence
30	DeepSeek V4 Pro 0813	85.3%	estimated ± 7.5 pp, low confidence
31	Claude Opus 4.7 (Adaptive)	85.3%	estimated ± 7.5 pp, low confidence
32	Claude Opus 4.8	85.2%	estimated ± 7.5 pp, low confidence
33	Grok 4.6	85.0%	estimated ± 7.5 pp, low confidence
34	Hy4 preview	84.8%	estimated ± 9.1 pp, low confidence
35	Kimi K3	84.7%	estimated ± 7.5 pp, low confidence
36	Grok 4.7	84.7%	estimated ± 7.5 pp, low confidence
37	Claude Opus 4.6 (Adaptive)	84.5%	estimated ± 7.5 pp, low confidence
38	GPT-5.6 Terra	84.4%	estimated ± 7.5 pp, low confidence
39	Claude Opus 4.5 Thinking	84.3%	estimated ± 7.5 pp, low confidence
40	DeepSeek V4.1 Flash	84.2%	estimated ± 7.5 pp, low confidence
41	Gemini 3 Flash	84.0%	estimated ± 7.5 pp, medium confidence
42	dots3-note Preview	83.8%	estimated ± 9.1 pp, medium confidence
43	Muse Spark 1.2	83.8%	estimated ± 7.5 pp, medium confidence
44	Claude Opus 4.7	83.5%	estimated ± 7.5 pp, medium confidence
45	Sakana Fugu	83.5%	estimated ± 9.2 pp, low confidence
46	Sakana Fugu-Ultra	83.5%	estimated ± 9.2 pp, low confidence
47	GPT-5.2	83.3%	estimated ± 7.5 pp, medium confidence
48	GPT-6 Luna	83.1%	estimated ± 7.5 pp, medium confidence
49	Muse Spark 1.3	83.0%	estimated ± 7.5 pp, medium confidence
50	GPT-5.6 Luna	82.5%	estimated ± 7.5 pp, medium confidence
51	Inkling	82.0%	estimated ± 7.5 pp, medium confidence
52	Step 5 Preview	81.9%	estimated ± 7.5 pp, medium confidence
53	Agents-A1	81.8%	estimated ± 9.1 pp, medium confidence
54	GPT-5.2-Codex	81.7%	estimated ± 7.5 pp, medium confidence
55	Grok 4	81.4%	estimated ± 7.5 pp, medium confidence
56	DeepSeek V4 Flash 0731	81.4%	estimated ± 7.5 pp, medium confidence
57	GPT-5 (high)	81.3%	estimated ± 7.5 pp, medium confidence
58	Claude Sonnet 5	81.2%	estimated ± 7.5 pp, medium confidence
59	GPT-5.1-Codex	81.1%	estimated ± 7.5 pp, medium confidence
60	GPT-5.1-Codex-Max	81.1%	estimated ± 7.5 pp, medium confidence
61	Kimi K2.7 Code	80.9%	estimated ± 7.5 pp, medium confidence
62	GPT-5 (medium)	80.9%	estimated ± 7.5 pp, medium confidence
63	Gemini 2.5 Pro	80.7%	estimated ± 7.5 pp, medium confidence
64	Ornith-1.5-397B	80.5%	estimated ± 9.1 pp, medium confidence
65	o3	80.4%	estimated ± 7.5 pp, medium confidence
66	Pareto 26.10 Preview	80.2%	estimated ± 13.9 pp, low confidence
67	Qwen3.8 Max	80.0%	estimated ± 9.1 pp, medium confidence
68	GPT-5.1	79.9%	estimated ± 7.5 pp, medium confidence
69	GPT-5.4 mini	79.7%	estimated ± 7.5 pp, medium confidence
70	Kimi K2.5 (Reasoning)	78.3%	estimated ± 7.5 pp, medium confidence
71	MiMo-V2.6-Pro	78.1%	estimated ± 7.5 pp, medium confidence
72	Grok 4.3	77.9%	estimated ± 7.5 pp, medium confidence
73	o1	77.9%	estimated ± 7.5 pp, medium confidence
74	GLM-5.3	77.5%	estimated ± 7.5 pp, medium confidence
75	Beam	77.2%	estimated ± 13.9 pp, low confidence
76	Inkling-Small	77.0%	estimated ± 7.5 pp, medium confidence
77	Kimi K2.6	76.6%	estimated ± 7.5 pp, medium confidence
78	Hy3	76.1%	estimated ± 7.5 pp, medium confidence
79	Qwen3.8-Omni-Flash	76.0%	estimated ± 9.1 pp, medium confidence
80	Apodex 1.1	75.9%	estimated ± 7.5 pp, medium confidence
81	Apodex 1.1 Mini	75.9%	estimated ± 7.5 pp, medium confidence
82	Qwen3.8 Max Preview	75.9%	estimated ± 7.5 pp, medium confidence
83	Hy3 Preview	75.7%	estimated ± 7.5 pp, medium confidence
84	Interfaze Beta	75.4%	estimated ± 9.2 pp, low confidence
85	DeepSeek-R1	75.0%	estimated ± 7.5 pp, medium confidence
86	Gemini 3.5 Flash-Lite	74.2%	estimated ± 7.5 pp, medium confidence
87	GLM-4.7	74.0%	estimated ± 7.5 pp, medium confidence
88	GLM-5V-Turbo	74.0%	estimated ± 7.5 pp, medium confidence
89	Grok 4.20	73.9%	estimated ± 13.9 pp, low confidence
90	Qwen 3.6 Max (preview)	73.9%	measured
91	Ling 3.1 Flash	73.8%	estimated ± 7.5 pp, medium confidence
92	DeepSeek V3.1 (Reasoning)	73.7%	estimated ± 7.5 pp, medium confidence
93	Qwen3.7 Max	73.6%	measured
94	GLM-5-Turbo	73.2%	estimated ± 7.5 pp, medium confidence
95	GPT-4.1	72.7%	estimated ± 7.5 pp, medium confidence
96	Kimi K2	72.3%	estimated ± 7.5 pp, medium confidence
97	MiMo-V2.6-Flash	72.0%	estimated ± 7.5 pp, medium confidence
98	Muse Glimmer 30B	72.0%	estimated ± 7.5 pp, medium confidence
99	MiniMax M2.7	71.8%	estimated ± 7.5 pp, medium confidence
100	MiMo-V2-Pro	71.6%	estimated ± 7.5 pp, medium confidence
101	Qwen3.6 Plus	71.6%	measured
102	Qwen3.7 Plus	71.4%	measured
103	Gemini 2.5 Flash	71.1%	estimated ± 7.5 pp, medium confidence
104	Claude 4.1 Opus Thinking	71.0%	estimated ± 10.4 pp, low confidence
105	Mistral Large 4	70.8%	estimated ± 7.5 pp, medium confidence
106	Step 3.7 Flash	70.8%	estimated ± 7.5 pp, medium confidence
107	Seed 2.1 Pro	70.8%	measured
108	GPT-5.4 nano	70.7%	estimated ± 7.5 pp, medium confidence
109	Claude Opus 4.5	70.6%	measured
110	Solar Open 2	70.6%	estimated ± 12.7 pp, low confidence
111	DeepSeek V3	70.5%	estimated ± 7.5 pp, medium confidence
112	Qwen3.5 397B	70.4%	measured
113	Grok 4.1 Fast (Reasoning)	70.1%	estimated ± 7.5 pp, medium confidence
114	Mistral Large 3	70.0%	estimated ± 7.5 pp, medium confidence
115	Llama 4 Maverick	69.9%	estimated ± 7.5 pp, medium confidence
116	Mistral Medium 3.5 128B	69.7%	estimated ± 7.5 pp, medium confidence
117	Qwen3.5 397B (Reasoning)	69.5%	estimated ± 7.5 pp, medium confidence
118	Qwen3.8-Flash-Next	69.5%	estimated ± 7.5 pp, medium confidence
119	o3-pro	69.5%	estimated ± 10.4 pp, low confidence
120	Qwen3 Max	69.4%	estimated ± 7.5 pp, medium confidence
121	DeepSeek V3 0324	69.3%	estimated ± 7.5 pp, medium confidence
122	GLM-5.2	69.3%	estimated ± 7.5 pp, medium confidence
123	Nemotron 3 Super 100B	69.3%	estimated ± 7.5 pp, medium confidence
124	Kimi K2.5	69.2%	measured
125	DeepSeek V3.2	69.0%	estimated ± 7.5 pp, medium confidence
126	GLM-5.1	68.7%	estimated ± 7.5 pp, medium confidence
127	Grok Code Fast 1	68.4%	estimated ± 7.5 pp, medium confidence
128	Llama 3.1 405B	68.1%	estimated ± 7.5 pp, medium confidence
129	DeepSeek V3.1	68.0%	estimated ± 7.5 pp, medium confidence
130	Grok 4 Fast (Reasoning)	67.6%	estimated ± 7.5 pp, medium confidence
131	Claude 4 Sonnet	67.5%	estimated ± 7.5 pp, medium confidence
132	Seed 2.1 Turbo	67.4%	measured
133	Ornith-1.5-35B-A3B	67.3%	estimated ± 9.1 pp, medium confidence
134	Trinity-Large-Preview	67.3%	estimated ± 7.5 pp, medium confidence
135	Trinity-Large-Thinking	67.3%	estimated ± 7.5 pp, medium confidence
136	MiMo-V2.5-Pro	67.2%	estimated ± 7.5 pp, medium confidence
137	Qwen3.5-122B-A10B	67.1%	measured
138	GLM-5	66.8%	measured
139	Mercury 2.5	66.7%	estimated ± 7.5 pp, medium confidence
140	MAI-Thinking-1	66.5%	estimated ± 9.2 pp, low confidence
141	GPT-OSS 120B	66.5%	estimated ± 7.5 pp, medium confidence
142	Mistral Small 4	66.3%	estimated ± 7.5 pp, medium confidence
143	Mistral Small 4 (Reasoning)	66.3%	estimated ± 7.5 pp, medium confidence
144	Nemotron 3 Ultra	66.2%	estimated ± 7.5 pp, medium confidence
145	Qwen3.6-27B	66.0%	measured
146	GLM-4.6	66.0%	estimated ± 7.5 pp, medium confidence
147	Qwen3.5-27B	65.6%	measured
148	Claude Sonnet 4.5	65.3%	estimated ± 9.2 pp, low confidence
149	Qwen3.6-35B-A3B	64.7%	measured
150	GPT-4.1 mini	64.6%	estimated ± 7.5 pp, medium confidence
151	Nemotron Ultra 253B	64.3%	estimated ± 7.5 pp, medium confidence
152	Gemma 4 31B	64.2%	estimated ± 7.5 pp, medium confidence
153	GPT-4o	64.0%	estimated ± 7.5 pp, medium confidence
154	Mistral Large 2	64.0%	estimated ± 7.5 pp, medium confidence
155	Claude 4.1 Opus	64.0%	estimated ± 10.4 pp, low confidence
156	Qwen3.5-35B-A3B	63.4%	measured
157	MiMo-V2-Omni	63.2%	estimated ± 7.5 pp, medium confidence
158	Gemma 4 26B A4B	62.9%	estimated ± 7.5 pp, medium confidence
159	Ultravox v0.6 Llama 3.3 70B	62.8%	estimated ± 7.5 pp, medium confidence
160	North Mini Code	62.6%	estimated ± 7.5 pp, medium confidence
161	Solar Pro 4	62.6%	estimated ± 7.5 pp, medium confidence
162	Qwen3 235B 2507	62.6%	measured
163	A.X K2	62.2%	estimated ± 7.5 pp, medium confidence
164	Solar Pro 3	62.1%	estimated ± 7.5 pp, medium confidence
165	Mistral Medium 3	61.8%	estimated ± 7.5 pp, medium confidence
166	Ling 3.0 Flash	61.6%	estimated ± 7.5 pp, medium confidence
167	Ling 3.0 Flash FP8	61.6%	estimated ± 7.5 pp, medium confidence
168	Ornith-1.5-9B	61.1%	estimated ± 9.1 pp, medium confidence
169	Claude 3 Haiku	60.7%	estimated ± 7.5 pp, medium confidence
170	Sarvam 105B	60.7%	estimated ± 7.5 pp, medium confidence
171	Nemotron 3 Nano 30B	60.2%	estimated ± 7.5 pp, medium confidence
172	Grok 4.1 Fast	60.1%	estimated ± 7.5 pp, medium confidence
173	Nova Pro	59.6%	estimated ± 7.5 pp, medium confidence
174	MiniMax M3	59.3%	estimated ± 7.5 pp, medium confidence
175	K-Exaone	58.8%	estimated ± 7.5 pp, medium confidence
176	GLM-4.5-Air	58.6%	estimated ± 7.5 pp, medium confidence
177	o1-pro	58.5%	estimated ± 9.2 pp, low confidence
178	Solar Pro 2	58.3%	estimated ± 7.5 pp, medium confidence
179	Ternary Bonsai 2 27B	58.2%	estimated ± 4.5 pp, high confidence
180	GPT-OSS 20B	58.1%	estimated ± 7.5 pp, medium confidence
181	Gemma 4 12B	57.4%	estimated ± 7.5 pp, medium confidence
182	Ling 2.6 Flash	57.4%	estimated ± 7.5 pp, medium confidence
183	MiMo-V2-Flash	57.4%	estimated ± 7.5 pp, medium confidence
184	Qwen3.8-27B	57.4%	estimated ± 7.5 pp, medium confidence
185	Quasar 438B	57.2%	estimated ± 7.5 pp, medium confidence
186	Llama 4 Scout	56.7%	estimated ± 7.5 pp, medium confidence
187	Nemotron 3 Nano Omni 30B A3B	56.7%	estimated ± 7.5 pp, medium confidence
188	LongCat-Flash-Lite-Sparse	56.1%	estimated ± 1.2 pp, low confidence
189	o3-mini	55.8%	estimated ± 9.2 pp, low confidence
190	Qwen3-Omni-30B-A3B-Thinking	55.6%	estimated ± 7.5 pp, medium confidence
191	Ling 3.0 Flash VL	55.2%	estimated ± 7.5 pp, medium confidence
192	Nemotron 3.5 Lightning 30B A3B NVFP4	55.2%	estimated ± 7.5 pp, medium confidence
193	Qwen3-Omni-30B-A3B-Instruct	55.0%	estimated ± 7.5 pp, medium confidence
194	Phi-4	54.6%	estimated ± 7.5 pp, medium confidence
195	GPT-4.1 nano	53.8%	estimated ± 7.5 pp, medium confidence
196	K-EXAONE 2.0	52.6%	estimated ± 7.5 pp, medium confidence
197	Gemma 3 27B	52.4%	estimated ± 7.5 pp, medium confidence
198	Mellum2-12B-A2.5B-Thinking	52.3%	estimated ± 4.5 pp, high confidence
199	Sarvam 30B	51.5%	estimated ± 7.5 pp, medium confidence
200	Granite 4.2 8B	48.3%	estimated ± 7.5 pp, medium confidence
201	o1-preview	48.1%	estimated ± 10.4 pp, low confidence
202	Celeris-1	47.8%	estimated ± 7.5 pp, medium confidence
203	ZAYA1-8B	47.4%	estimated ± 9.2 pp, low confidence
204	Exaone 4.0 32B	46.8%	estimated ± 7.5 pp, medium confidence
205	Granite 4.2 30B	45.5%	estimated ± 7.5 pp, medium confidence
206	LFM2.5-8B-A1B	43.6%	estimated ± 7.5 pp, medium confidence
207	Granite 4.2 3B	43.0%	estimated ± 7.5 pp, medium confidence
208	Command A+	42.2%	estimated ± 7.5 pp, medium confidence
209	Gemma 4 E4B	41.3%	estimated ± 7.5 pp, medium confidence
210	Ling 3.0 Tiny	41.0%	estimated ± 7.5 pp, medium confidence
211	MiniCPM5-2B	40.8%	measured
212	Claude 3 Opus	40.2%	estimated ± 10.4 pp, low confidence
213	DeepSeek R1 Distill Qwen 32B	39.1%	estimated ± 10.4 pp, low confidence
214	Gemini 1.5 Pro	37.5%	estimated ± 10.4 pp, low confidence
215	Claude 3.5 Sonnet	36.4%	estimated ± 9.2 pp, low confidence
216	Mellum2-12B-A2.5B-Instruct	35.9%	estimated ± 4.5 pp, high confidence
217	ZAYA1-74B-Preview	35.1%	estimated ± 9.2 pp, low confidence
218	Gemma 4 E2B	34.7%	estimated ± 7.5 pp, low confidence
219	GPT-4 Turbo	34.4%	estimated ± 10.4 pp, low confidence
220	Qwen2.5 Coder 32B Instruct	33.3%	estimated ± 10.4 pp, low confidence
221	GPT-4o mini	33.0%	estimated ± 10.4 pp, low confidence
222	LFM2.5-VL-1.6B-Extract	31.7%	estimated ± 7.5 pp, low confidence
223	Soofi S 30B-A3B	30.4%	estimated ± 9.2 pp, low confidence
224	Phi-4 Multimodal Instruct	29.8%	estimated ± 10.4 pp, low confidence
225	Granite-4.0-H-1B	29.3%	estimated ± 7.5 pp, low confidence
226	LFM2.5-VL-450M	29.3%	estimated ± 9.2 pp, low confidence
227	LFM2.5-230M	29.3%	estimated ± 9.2 pp, low confidence
228	Exaone 4.0 1.2B	28.4%	estimated ± 7.5 pp, low confidence
229	Gemini 1.0 Pro	27.9%	estimated ± 10.4 pp, low confidence
230	LFM2.5-2.6B	25.8%	estimated ± 7.5 pp, low confidence
231	LLaDA2.2-mini	23.7%	estimated ± 13.9 pp, low confidence
232	Granite-4.0-350M	23.5%	estimated ± 7.5 pp, low confidence
233	MiniCPM5-1B	23.1%	measured
234	Granite-4.0-H-350M	23.0%	estimated ± 7.5 pp, low confidence

SuperGPQA leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General