Knowledge & reasoning

MMLU leaderboard

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

As of 2026-10-07, the highest measured score on MMLU is 91.8% by o1. 215 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	Claude Haiku 5.5	100.0%	estimated ± 1.9 pp, low confidence
2	GLM-5.3-Flash	100.0%	estimated ± 1.9 pp, low confidence
3	Claude Fable 5.1	96.6%	estimated ± 1.1 pp, low confidence
4	Claude Opus 5.5	96.5%	estimated ± 1.1 pp, low confidence
5	Claude Fable 5	96.5%	estimated ± 1.1 pp, low confidence
6	Claude 4.1 Opus Thinking	96.4%	estimated ± 1.9 pp, low confidence
7	GPT-6 Astra	96.3%	estimated ± 1.1 pp, low confidence
8	GPT-6.1 Sol	96.2%	estimated ± 1.1 pp, low confidence
9	Claude Opus 5	96.1%	estimated ± 1.1 pp, low confidence
10	o3-pro	96.0%	estimated ± 1.9 pp, low confidence
11	GPT-5.6 Sol	96.0%	estimated ± 1.1 pp, low confidence
12	GPT-5.5	95.8%	estimated ± 1.1 pp, low confidence
13	Gemini 3 Pro	95.6%	estimated ± 1.1 pp, low confidence
14	Gemini 3.7 Flash	95.6%	estimated ± 1.1 pp, low confidence
15	Gemini 3.1 Pro	95.5%	estimated ± 1.1 pp, low confidence
16	Gemini 3.8 Flash	95.5%	estimated ± 1.1 pp, low confidence
17	GPT-6 Sol	95.5%	estimated ± 1.1 pp, low confidence
18	Claude Sonnet 5.5	95.4%	estimated ± 1.1 pp, low confidence
19	GPT-5.3 Codex	95.3%	estimated ± 1.1 pp, low confidence
20	Muse Spark 1.1	95.2%	estimated ± 1.1 pp, low confidence
21	Gemini 3.5 Flash	95.2%	estimated ± 1.1 pp, low confidence
22	Grok 4.5	95.1%	estimated ± 1.1 pp, low confidence
23	GPT-5.4	95.0%	estimated ± 1.1 pp, low confidence
24	Gemini 3.6 Flash	94.9%	estimated ± 1.1 pp, low confidence
25	Gemini 4 Argon	94.9%	estimated ± 1.1 pp, low confidence
26	Muse Spark	94.9%	estimated ± 1.1 pp, low confidence
27	DeepSeek V4 Pro 0813	94.8%	estimated ± 1.1 pp, low confidence
28	Claude Opus 4.7 (Adaptive)	94.8%	estimated ± 1.1 pp, low confidence
29	Claude Opus 4.8	94.8%	estimated ± 1.1 pp, low confidence
30	Grok 4.6	94.7%	estimated ± 1.1 pp, low confidence
31	Kimi K3	94.6%	estimated ± 1.1 pp, low confidence
32	Grok 4.7	94.6%	estimated ± 1.1 pp, low confidence
33	Claude Opus 4.6 (Adaptive)	94.5%	estimated ± 1.1 pp, low confidence
34	GPT-5.6 Terra	94.5%	estimated ± 1.1 pp, low confidence
35	Claude Opus 4.5 Thinking	94.5%	estimated ± 1.1 pp, low confidence
36	DeepSeek V4.1 Flash	94.4%	estimated ± 1.1 pp, low confidence
37	Claude Opus 4.6	94.4%	estimated ± 1.1 pp, low confidence
38	Gemini 3 Flash	94.4%	estimated ± 1.1 pp, low confidence
39	Muse Spark 1.2	94.3%	estimated ± 1.1 pp, low confidence
40	Claude 4.1 Opus	94.2%	estimated ± 1.9 pp, low confidence
41	Claude Opus 4.7	94.2%	estimated ± 1.1 pp, low confidence
42	GPT-5.2	94.1%	estimated ± 1.1 pp, low confidence
43	GPT-6 Luna	94.0%	estimated ± 1.1 pp, low confidence
44	Muse Spark 1.3	94.0%	estimated ± 1.1 pp, low confidence
45	GPT-5.6 Luna	93.9%	estimated ± 1.1 pp, low confidence
46	Inkling	93.7%	estimated ± 1.1 pp, low confidence
47	Step 5 Preview	93.6%	estimated ± 1.1 pp, low confidence
48	GPT-5.2-Codex	93.6%	estimated ± 1.1 pp, low confidence
49	Claude Opus 4.5	93.5%	estimated ± 1.1 pp, low confidence
50	Grok 4	93.5%	estimated ± 1.1 pp, low confidence
51	DeepSeek V4 Flash 0731	93.4%	estimated ± 1.1 pp, low confidence
52	GPT-5 (high)	93.4%	estimated ± 1.1 pp, low confidence
53	Claude Sonnet 5	93.4%	estimated ± 1.1 pp, low confidence
54	GPT-5.1-Codex	93.3%	estimated ± 1.1 pp, low confidence
55	GPT-5.1-Codex-Max	93.3%	estimated ± 1.1 pp, low confidence
56	Kimi K2.7 Code	93.3%	estimated ± 1.1 pp, low confidence
57	GPT-5 (medium)	93.3%	estimated ± 1.1 pp, low confidence
58	Gemini 2.5 Pro	93.2%	estimated ± 1.1 pp, low confidence
59	Claude Sonnet 4.6	93.1%	estimated ± 1.1 pp, low confidence
60	o3	93.1%	estimated ± 1.1 pp, low confidence
61	Qwen 3.6 Max (preview)	92.9%	estimated ± 1.1 pp, low confidence
62	GPT-5.1	92.9%	estimated ± 1.1 pp, low confidence
63	GPT-5.4 mini	92.8%	estimated ± 1.1 pp, low confidence
64	Kimi K2.5	92.3%	estimated ± 1.1 pp, low confidence
65	Kimi K2.5 (Reasoning)	92.3%	estimated ± 1.1 pp, low confidence
66	MiMo-V2.6-Pro	92.2%	estimated ± 1.1 pp, low confidence
67	Grok 4.3	92.2%	estimated ± 1.1 pp, low confidence
68	GLM-5.3	92.0%	estimated ± 1.1 pp, medium confidence
69	o1	91.8%	measured
70	Inkling-Small	91.8%	estimated ± 1.1 pp, medium confidence
71	Kimi K2.6	91.6%	estimated ± 1.1 pp, medium confidence
72	Hy3	91.5%	estimated ± 1.1 pp, medium confidence
73	Apodex 1.1	91.4%	estimated ± 1.1 pp, medium confidence
74	Apodex 1.1 Mini	91.4%	estimated ± 1.1 pp, medium confidence
75	Qwen3.8 Max Preview	91.4%	estimated ± 1.1 pp, medium confidence
76	Hy3 Preview	91.3%	estimated ± 1.1 pp, medium confidence
77	Qwen3.7 Max	91.2%	estimated ± 1.1 pp, medium confidence
78	DeepSeek-R1	91.0%	estimated ± 1.1 pp, medium confidence
79	Gemini 3.5 Flash-Lite	90.7%	estimated ± 1.1 pp, medium confidence
80	GLM-4.7	90.6%	estimated ± 1.1 pp, medium confidence
81	GLM-5V-Turbo	90.6%	estimated ± 1.1 pp, medium confidence
82	Ling 3.1 Flash	90.5%	estimated ± 1.1 pp, medium confidence
83	DeepSeek V3.1 (Reasoning)	90.5%	estimated ± 1.1 pp, medium confidence
84	Sakana Fugu	90.5%	estimated ± 3.6 pp, low confidence
85	Sakana Fugu-Ultra	90.5%	estimated ± 3.6 pp, low confidence
86	Claude Mythos 5	90.5%	estimated ± 3.6 pp, low confidence
87	Ornith-1.5-397B	90.5%	estimated ± 3.6 pp, low confidence
88	Qwen3.8 Max	90.5%	estimated ± 3.6 pp, low confidence
89	Hy4 preview	90.4%	estimated ± 3.6 pp, low confidence
90	Qwen3.8-Omni-Flash	90.4%	estimated ± 3.6 pp, low confidence
91	Interfaze Beta	90.4%	estimated ± 3.6 pp, low confidence
92	Ornith-1.5-35B-A3B	90.4%	estimated ± 3.6 pp, low confidence
93	Ornith-1.5-9B	90.3%	estimated ± 3.6 pp, low confidence
94	GLM-5-Turbo	90.3%	estimated ± 1.1 pp, medium confidence
95	Ternary Bonsai 2 27B	90.3%	estimated ± 3.6 pp, low confidence
96	MAI-Thinking-1	90.2%	estimated ± 3.6 pp, low confidence
97	GPT-4.1	90.2%	measured
98	Claude Sonnet 4.5	90.2%	estimated ± 3.6 pp, low confidence
99	Kimi K2	89.9%	estimated ± 1.1 pp, medium confidence
100	Qwen3 235B 2507	89.9%	estimated ± 3.6 pp, low confidence
101	MiMo-V2.6-Flash	89.8%	estimated ± 1.1 pp, medium confidence
102	Muse Glimmer 30B	89.8%	estimated ± 1.1 pp, medium confidence
103	MiniMax M2.7	89.7%	estimated ± 1.1 pp, medium confidence
104	MiMo-V2-Pro	89.6%	estimated ± 1.1 pp, medium confidence
105	Qwen3.6 Plus	89.5%	estimated ± 1.1 pp, medium confidence
106	GLM-5	89.5%	estimated ± 1.1 pp, medium confidence
107	Gemini 2.5 Flash	89.4%	estimated ± 1.1 pp, medium confidence
108	ZAYA1-8B	89.3%	estimated ± 3.6 pp, medium confidence
109	Mistral Large 4	89.3%	estimated ± 1.1 pp, medium confidence
110	Step 3.7 Flash	89.3%	estimated ± 1.1 pp, medium confidence
111	GPT-5.4 nano	89.2%	estimated ± 1.1 pp, medium confidence
112	DeepSeek V3	89.1%	estimated ± 1.1 pp, medium confidence
113	o1-pro	89.0%	estimated ± 1.9 pp, medium confidence
114	Grok 4.1 Fast (Reasoning)	89.0%	estimated ± 1.1 pp, medium confidence
115	Mistral Large 3	88.9%	estimated ± 1.1 pp, medium confidence
116	Llama 4 Maverick	88.9%	estimated ± 1.1 pp, medium confidence
117	Mistral Medium 3.5 128B	88.8%	estimated ± 1.1 pp, medium confidence
118	Qwen3.5 397B	88.7%	estimated ± 1.1 pp, medium confidence
119	Qwen3.5 397B (Reasoning)	88.7%	estimated ± 1.1 pp, medium confidence
120	Qwen3.8-Flash-Next	88.7%	estimated ± 1.1 pp, medium confidence
121	Qwen3.5-122B-A10B	88.6%	estimated ± 1.1 pp, medium confidence
122	Qwen3 Max	88.6%	estimated ± 1.1 pp, medium confidence
123	DeepSeek V3 0324	88.6%	estimated ± 1.1 pp, medium confidence
124	GLM-5.2	88.6%	estimated ± 1.1 pp, medium confidence
125	Nemotron 3 Super 100B	88.6%	estimated ± 1.1 pp, medium confidence
126	DeepSeek V3.2	88.5%	estimated ± 1.1 pp, medium confidence
127	GLM-5.1	88.3%	estimated ± 1.1 pp, medium confidence
128	Grok Code Fast 1	88.2%	estimated ± 1.1 pp, medium confidence
129	Llama 3.1 405B	88.1%	estimated ± 1.1 pp, medium confidence
130	DeepSeek V3.1	88.0%	estimated ± 1.1 pp, medium confidence
131	Grok 4 Fast (Reasoning)	87.9%	estimated ± 1.1 pp, medium confidence
132	Claude 4 Sonnet	87.8%	estimated ± 1.1 pp, medium confidence
133	Qwen3.7 Plus	87.7%	estimated ± 1.1 pp, medium confidence
134	Trinity-Large-Thinking	87.7%	estimated ± 1.1 pp, medium confidence
135	MiMo-V2.5-Pro	87.6%	estimated ± 1.1 pp, medium confidence
136	o1-preview	87.6%	estimated ± 1.9 pp, medium confidence
137	GPT-4.1 mini	87.5%	measured
138	Mercury 2.5	87.4%	estimated ± 1.1 pp, medium confidence
139	GPT-OSS 120B	87.3%	estimated ± 1.1 pp, medium confidence
140	Mistral Small 4	87.2%	estimated ± 1.1 pp, medium confidence
141	Mistral Small 4 (Reasoning)	87.2%	estimated ± 1.1 pp, medium confidence
142	Trinity-Large-Preview	87.2%	measured
143	Nemotron 3 Ultra	87.2%	estimated ± 1.1 pp, medium confidence
144	GLM-4.6	87.1%	estimated ± 1.1 pp, medium confidence
145	o3-mini	86.9%	measured
146	Qwen3.5-27B	86.6%	estimated ± 1.1 pp, medium confidence
147	Claude 3.5 Sonnet	86.6%	estimated ± 3.6 pp, medium confidence
148	Nemotron Ultra 253B	86.3%	estimated ± 1.1 pp, medium confidence
149	Qwen3.5-35B-A3B	86.3%	estimated ± 1.1 pp, medium confidence
150	Gemma 4 31B	86.2%	estimated ± 1.1 pp, medium confidence
151	GPT-4o	86.1%	estimated ± 1.1 pp, medium confidence
152	Mistral Large 2	86.1%	estimated ± 1.1 pp, medium confidence
153	Qwen3.6-27B	85.9%	estimated ± 1.1 pp, medium confidence
154	Mellum2-12B-A2.5B-Thinking	85.8%	estimated ± 3.6 pp, medium confidence
155	MiMo-V2-Omni	85.7%	estimated ± 1.1 pp, medium confidence
156	ZAYA1-74B-Preview	85.6%	estimated ± 3.6 pp, medium confidence
157	Gemma 4 26B A4B	85.6%	estimated ± 1.1 pp, medium confidence
158	Ultravox v0.6 Llama 3.3 70B	85.5%	estimated ± 1.1 pp, medium confidence
159	North Mini Code	85.4%	estimated ± 1.1 pp, medium confidence
160	Solar Pro 4	85.4%	estimated ± 1.1 pp, medium confidence
161	Qwen3.6-35B-A3B	85.4%	estimated ± 1.1 pp, medium confidence
162	LongCat-Flash-Lite-Sparse	85.3%	measured
163	A.X K2	85.2%	estimated ± 1.1 pp, medium confidence
164	Solar Pro 3	85.1%	estimated ± 1.1 pp, medium confidence
165	Mistral Medium 3	85.0%	estimated ± 1.1 pp, medium confidence
166	Ling 3.0 Flash	84.9%	estimated ± 1.1 pp, medium confidence
167	Ling 3.0 Flash FP8	84.9%	estimated ± 1.1 pp, medium confidence
168	Claude 3 Haiku	84.4%	estimated ± 1.1 pp, medium confidence
169	Sarvam 105B	84.4%	estimated ± 1.1 pp, medium confidence
170	Nemotron 3 Nano 30B	84.2%	estimated ± 1.1 pp, medium confidence
171	Grok 4.1 Fast	84.1%	estimated ± 1.1 pp, medium confidence
172	Nova Pro	83.8%	estimated ± 1.1 pp, medium confidence
173	MiniMax M3	83.7%	estimated ± 1.1 pp, medium confidence
174	K-Exaone	83.4%	estimated ± 1.1 pp, medium confidence
175	GLM-4.5-Air	83.3%	estimated ± 1.1 pp, medium confidence
176	Solar Pro 2	83.1%	estimated ± 1.1 pp, medium confidence
177	Claude 3 Opus	83.0%	estimated ± 1.9 pp, medium confidence
178	GPT-OSS 20B	83.0%	estimated ± 1.1 pp, medium confidence
179	Gemma 4 12B	82.6%	estimated ± 1.1 pp, medium confidence
180	Ling 2.6 Flash	82.6%	estimated ± 1.1 pp, medium confidence
181	MiMo-V2-Flash	82.6%	estimated ± 1.1 pp, medium confidence
182	Qwen3.8-27B	82.6%	estimated ± 1.1 pp, medium confidence
183	Quasar 438B	82.5%	estimated ± 1.1 pp, medium confidence
184	DeepSeek R1 Distill Qwen 32B	82.3%	estimated ± 1.9 pp, medium confidence
185	Llama 4 Scout	82.2%	estimated ± 1.1 pp, medium confidence
186	Nemotron 3 Nano Omni 30B A3B	82.2%	estimated ± 1.1 pp, medium confidence
187	Qwen3-Omni-30B-A3B-Thinking	81.6%	estimated ± 1.1 pp, medium confidence
188	Ling 3.0 Flash VL	81.3%	estimated ± 1.1 pp, medium confidence
189	Nemotron 3.5 Lightning 30B A3B NVFP4	81.3%	estimated ± 1.1 pp, medium confidence
190	Qwen3-Omni-30B-A3B-Instruct	81.2%	estimated ± 1.1 pp, medium confidence
191	Gemini 1.5 Pro	81.2%	estimated ± 1.9 pp, medium confidence
192	Phi-4	81.0%	estimated ± 1.1 pp, medium confidence
193	GPT-4.1 nano	80.1%	measured
194	K-EXAONE 2.0	79.7%	estimated ± 1.1 pp, low confidence
195	Gemma 3 27B	79.6%	estimated ± 1.1 pp, low confidence
196	Sarvam 30B	79.1%	estimated ± 1.1 pp, low confidence
197	GPT-4 Turbo	78.8%	estimated ± 1.9 pp, low confidence
198	Qwen2.5 Coder 32B Instruct	77.9%	estimated ± 1.9 pp, low confidence
199	GPT-4o mini	77.6%	estimated ± 1.9 pp, low confidence
200	Granite 4.2 8B	76.9%	estimated ± 1.1 pp, low confidence
201	Celeris-1	76.6%	estimated ± 1.1 pp, low confidence
202	Exaone 4.0 32B	75.9%	estimated ± 1.1 pp, low confidence
203	Granite 4.2 30B	74.9%	estimated ± 1.1 pp, low confidence
204	Phi-4 Multimodal Instruct	74.6%	estimated ± 1.9 pp, low confidence
205	LFM2.5-8B-A1B	73.5%	estimated ± 1.1 pp, low confidence
206	Granite 4.2 3B	73.0%	estimated ± 1.1 pp, low confidence
207	Gemini 1.0 Pro	72.7%	estimated ± 1.9 pp, low confidence
208	Command A+	72.3%	estimated ± 1.1 pp, low confidence
209	Gemma 4 E4B	71.6%	estimated ± 1.1 pp, low confidence
210	Ling 3.0 Tiny	71.4%	estimated ± 1.1 pp, low confidence
211	MiniCPM5-2B	71.1%	estimated ± 1.1 pp, low confidence
212	Soofi S 30B-A3B	68.9%	estimated ± 3.6 pp, low confidence
213	Gemma 4 E2B	65.7%	estimated ± 1.1 pp, low confidence
214	LFM2.5-VL-1.6B-Extract	62.6%	estimated ± 1.1 pp, low confidence
215	Mellum2-12B-A2.5B-Instruct	62.5%	estimated ± 3.6 pp, low confidence
216	Granite-4.0-H-1B	60.0%	estimated ± 1.1 pp, low confidence
217	Exaone 4.0 1.2B	59.0%	estimated ± 1.1 pp, low confidence
218	LFM2.5-2.6B	55.8%	estimated ± 1.1 pp, low confidence
219	Granite-4.0-350M	52.8%	estimated ± 1.1 pp, low confidence
220	Granite-4.0-H-350M	52.1%	estimated ± 1.1 pp, low confidence
221	LFM2.5-VL-450M	10.8%	estimated ± 3.6 pp, low confidence
222	LFM2.5-230M	10.2%	estimated ± 3.6 pp, low confidence

MMLU leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General