Knowledge & reasoning

Vals MMLU-Pro leaderboard

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

As of 2026-10-07, the highest measured score on Vals MMLU-Pro is 92.4% by Claude Fable 5.1. 186 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	Claude Sonnet 5.5	96.7%	estimated ± 1.5 pp, low confidence
2	Claude Opus 5.5	94.0%	estimated ± 1.5 pp, low confidence
3	Ling 3.1 Flash	93.8%	estimated ± 1.5 pp, low confidence
4	GPT-6.1 Sol	93.0%	estimated ± 1.5 pp, low confidence
5	Gemini 4 Argon	92.8%	estimated ± 1.4 pp, low confidence
6	Claude Fable 5.1	92.4%	measured
7	Claude Opus 5	91.6%	measured
8	Claude Fable 5	91.5%	measured
9	Gemini 3.1 Pro	91.0%	measured
10	GPT-6 Luna	90.3%	estimated ± 1.5 pp, low confidence
11	GPT-6 Sol	90.3%	estimated ± 1.5 pp, low confidence
12	Gemini 3.8 Flash	90.2%	measured
13	Gemini 3.7 Flash	90.1%	measured
14	Claude Opus 4.7	89.9%	measured
15	dots3-note Preview	89.9%	estimated ± 2.0 pp, high confidence
16	Pareto 26.9	89.7%	estimated ± 1.9 pp, high confidence
17	Claude Opus 4.8	89.6%	measured
18	Gemini 3.5 Flash	89.5%	measured
19	Grok 4.6	89.4%	measured
20	GPT-5.5 Pro	89.3%	estimated ± 1.8 pp, high confidence
21	Gemini 3.6 Flash	89.3%	measured
22	Qwen3.7 Max	89.3%	measured
23	GPT-5.4 Pro	89.2%	estimated ± 1.8 pp, high confidence
24	GPT-5.3 Codex	89.2%	estimated ± 2.2 pp, high confidence
25	Grok 4.5	89.2%	measured
26	Claude Haiku 5.5	89.1%	estimated ± 1.9 pp, high confidence
27	GPT-5.6 Sol	89.1%	measured
28	Claude Opus 4.6 (Adaptive)	89.0%	estimated ± 1.8 pp, high confidence
29	GPT-6 Astra	88.9%	estimated ± 1.2 pp, medium confidence
30	Sakana Fugu	88.8%	estimated ± 1.2 pp, medium confidence
31	Sakana Fugu-Ultra	88.8%	estimated ± 1.2 pp, medium confidence
32	Muse Spark 1.1	88.7%	measured
33	Gemini 3 Flash	88.6%	measured
34	Qwen3.8 Max	88.6%	measured
35	Claude Opus 4.7 (Adaptive)	88.5%	estimated ± 1.2 pp, high confidence
36	Step 5 Preview	88.5%	estimated ± 1.6 pp, high confidence
37	Claude Mythos 5	88.5%	estimated ± 1.2 pp, high confidence
38	Muse Spark 1.2	88.3%	measured
39	GPT-5.4	88.1%	estimated ± 1.2 pp, high confidence
40	Ornith-1.5-397B	88.1%	estimated ± 1.2 pp, high confidence
41	GPT-5.5	88.1%	measured
42	Kimi K3	88.0%	measured
43	GPT-5.2	88.0%	estimated ± 1.2 pp, high confidence
44	Hy4 preview	87.9%	estimated ± 1.2 pp, high confidence
45	Pareto 26.10 Preview	87.9%	estimated ± 1.6 pp, high confidence
46	Qwen3.8-Flash-Next	87.7%	estimated ± 1.2 pp, high confidence
47	Qwen3.6 Plus	87.7%	measured
48	Kimi K2.6	87.6%	measured
49	Claude Opus 4.6	87.5%	estimated ± 1.2 pp, high confidence
50	Muse Spark 1.3	87.5%	estimated ± 2.2 pp, high confidence
51	Claude Sonnet 5	87.5%	measured
52	Qwen3.8-Omni-Flash	87.4%	estimated ± 1.2 pp, high confidence
53	DeepSeek V4.1 Flash	87.4%	estimated ± 1.2 pp, high confidence
54	Claude Sonnet 4.6	87.3%	measured
55	Muse Spark	87.3%	measured
56	Agents-A1	87.2%	estimated ± 2.5 pp, medium confidence
57	Qwen3.7 Plus	87.0%	estimated ± 1.2 pp, high confidence
58	GPT-5.2-Codex	87.0%	estimated ± 2.2 pp, high confidence
59	DeepSeek V4 Pro 0813	87.0%	measured
60	Grok 4	86.9%	estimated ± 2.2 pp, high confidence
61	GLM-5.1	86.9%	measured
62	GPT-5 (high)	86.9%	estimated ± 2.2 pp, high confidence
63	Beam	86.9%	estimated ± 1.6 pp, high confidence
64	Grok 4.7	86.8%	estimated ± 1.5 pp, medium confidence
65	GPT-5.1-Codex	86.8%	estimated ± 2.2 pp, high confidence
66	GPT-5.1-Codex-Max	86.8%	estimated ± 2.2 pp, high confidence
67	Interfaze Beta	86.8%	estimated ± 1.2 pp, high confidence
68	GLM-5.3	86.8%	measured
69	Kimi K2.7 Code	86.8%	estimated ± 2.2 pp, high confidence
70	GPT-5 (medium)	86.7%	estimated ± 2.2 pp, high confidence
71	GLM-5.2	86.7%	measured
72	GPT-5.6 Terra	86.7%	measured
73	o3	86.6%	estimated ± 2.2 pp, high confidence
74	Claude Opus 4.5 Thinking	86.5%	estimated ± 1.8 pp, high confidence
75	Qwen 3.6 Max (preview)	86.4%	estimated ± 2.2 pp, high confidence
76	Ornith-1.5-35B-A3B	86.4%	estimated ± 1.2 pp, high confidence
77	Grok 4.20	86.3%	measured
78	Inkling	86.3%	measured
79	DeepSeek V4 Flash 0731	86.2%	measured
80	Gemini 3.1 Flash-Lite	86.2%	measured
81	GLM-5.3-Flash	86.1%	measured
82	Solar Pro 4	86.1%	estimated ± 1.6 pp, high confidence
83	GPT-5.6 Luna	86.0%	measured
84	MiMo-V2.6-Pro	85.9%	estimated ± 2.2 pp, high confidence
85	Gemini 3.5 Flash-Lite	85.8%	measured
86	Grok 4.3	85.8%	measured
87	Nemotron 3 Ultra	85.8%	measured
88	Qwen3.5 397B	85.8%	estimated ± 1.2 pp, high confidence
89	Gemini 3 Pro Deep Think	85.7%	estimated ± 2.0 pp, high confidence
90	Inkling-Small	85.6%	measured
91	Gemini 3 Pro	85.4%	estimated ± 1.8 pp, high confidence
92	Hy3	85.3%	estimated ± 2.2 pp, high confidence
93	Apodex 1.1	85.3%	estimated ± 2.2 pp, high confidence
94	Apodex 1.1 Mini	85.3%	estimated ± 2.2 pp, high confidence
95	Qwen3.8 Max Preview	85.3%	estimated ± 2.2 pp, high confidence
96	Qwen3.6-27B	85.2%	estimated ± 1.2 pp, high confidence
97	DeepSeek-R1	85.1%	estimated ± 2.2 pp, high confidence
98	Kimi K2.5	85.0%	estimated ± 1.2 pp, high confidence
99	Kimi K2.5 (Reasoning)	85.0%	estimated ± 1.2 pp, high confidence
100	GPT-5.1	84.9%	estimated ± 1.8 pp, high confidence
101	GLM-5V-Turbo	84.8%	estimated ± 2.2 pp, high confidence
102	DeepSeek V3.1 (Reasoning)	84.8%	estimated ± 2.2 pp, high confidence
103	GLM-5-Turbo	84.7%	estimated ± 2.2 pp, high confidence
104	Hy3 Preview	84.6%	estimated ± 1.2 pp, high confidence
105	GPT-5.4 mini	84.6%	measured
106	MiMo-V2.5-Pro	84.6%	measured
107	Solar Open 2	84.6%	estimated ± 1.6 pp, high confidence
108	Kimi K2	84.5%	estimated ± 2.2 pp, high confidence
109	Claude Opus 4.5	84.4%	estimated ± 1.2 pp, high confidence
110	MiMo-V2.6-Flash	84.4%	estimated ± 2.2 pp, high confidence
111	Muse Glimmer 30B	84.4%	estimated ± 2.2 pp, high confidence
112	MiMo-V2-Pro	84.3%	estimated ± 2.2 pp, high confidence
113	Qwen3.8-27B	84.3%	measured
114	Gemini 2.5 Flash	84.3%	estimated ± 2.2 pp, high confidence
115	o3-pro	84.2%	estimated ± 2.6 pp, medium confidence
116	MiniMax M3	84.2%	measured
117	Mistral Large 4	84.2%	estimated ± 2.2 pp, high confidence
118	Step 3.7 Flash	84.2%	estimated ± 2.2 pp, high confidence
119	A.X K2	84.2%	estimated ± 1.6 pp, high confidence
120	Qwen3.5 Flash	84.1%	measured
121	Grok 4.1 Fast (Reasoning)	84.1%	estimated ± 2.2 pp, high confidence
122	Mistral Large 3	84.1%	estimated ± 2.2 pp, high confidence
123	Llama 4 Maverick	84.0%	estimated ± 2.2 pp, high confidence
124	Qwen3.5-122B-A10B	84.0%	estimated ± 1.2 pp, high confidence
125	Qwen3.5 397B (Reasoning)	84.0%	estimated ± 2.2 pp, high confidence
126	Qwen3 Max	83.9%	estimated ± 2.2 pp, high confidence
127	DeepSeek V3 0324	83.9%	estimated ± 2.2 pp, high confidence
128	Nemotron 3 Super 100B	83.9%	estimated ± 2.2 pp, high confidence
129	DeepSeek V3.2	83.9%	estimated ± 2.2 pp, high confidence
130	Grok Code Fast 1	83.8%	estimated ± 2.2 pp, high confidence
131	Ornith-1.5-9B	83.7%	estimated ± 1.2 pp, high confidence
132	Llama 3.1 405B	83.7%	estimated ± 2.2 pp, high confidence
133	DeepSeek V3.1	83.7%	estimated ± 2.2 pp, high confidence
134	Grok 4 Fast (Reasoning)	83.6%	estimated ± 2.2 pp, high confidence
135	Claude 4 Sonnet	83.6%	estimated ± 2.2 pp, high confidence
136	GPT-OSS 120B	83.5%	estimated ± 2.2 pp, high confidence
137	Mistral Small 4	83.4%	estimated ± 2.2 pp, high confidence
138	Mistral Small 4 (Reasoning)	83.4%	estimated ± 2.2 pp, high confidence
139	GLM-5	83.2%	estimated ± 1.2 pp, high confidence
140	Qwen3.6-35B-A3B	83.2%	estimated ± 1.2 pp, high confidence
141	Claude 4.1 Opus Thinking	83.2%	estimated ± 2.5 pp, high confidence
142	Nemotron Ultra 253B	83.1%	estimated ± 2.2 pp, high confidence
143	Phi-4 Multimodal Instruct	83.1%	estimated ± 2.5 pp, medium confidence
144	DeepSeek R1 Distill Qwen 32B	83.1%	estimated ± 2.5 pp, medium confidence
145	Gemini 1.5 Pro	83.1%	estimated ± 2.5 pp, medium confidence
146	Gemini 1.0 Pro	83.1%	estimated ± 2.5 pp, medium confidence
147	GPT-4o mini	83.1%	estimated ± 2.5 pp, medium confidence
148	GPT-4o	83.1%	estimated ± 2.2 pp, high confidence
149	Mistral Large 2	83.1%	estimated ± 2.2 pp, high confidence
150	Qwen2.5 Coder 32B Instruct	83.1%	estimated ± 2.5 pp, medium confidence
151	GPT-4 Turbo	83.1%	estimated ± 2.5 pp, medium confidence
152	Claude 3 Opus	83.1%	estimated ± 2.5 pp, medium confidence
153	MiMo-V2-Omni	83.0%	estimated ± 2.2 pp, high confidence
154	Ultravox v0.6 Llama 3.3 70B	82.9%	estimated ± 2.2 pp, high confidence
155	North Mini Code	82.9%	estimated ± 2.2 pp, high confidence
156	Ternary Bonsai 2 27B	82.9%	estimated ± 1.2 pp, high confidence
157	MiMo-V2.5	82.9%	measured
158	Solar Pro 3	82.8%	estimated ± 2.2 pp, high confidence
159	Mistral Medium 3	82.8%	estimated ± 2.2 pp, high confidence
160	Claude 4.1 Opus	82.7%	estimated ± 2.6 pp, medium confidence
161	GLM-4.7	82.7%	measured
162	Claude 3 Haiku	82.7%	estimated ± 2.2 pp, high confidence
163	Sarvam 105B	82.7%	estimated ± 2.2 pp, high confidence
164	Nemotron 3 Nano 30B	82.6%	estimated ± 2.2 pp, high confidence
165	Grok 4.1 Fast	82.6%	estimated ± 2.2 pp, high confidence
166	Nova Pro	82.5%	estimated ± 2.2 pp, high confidence
167	Qwen3.5-27B	82.5%	estimated ± 1.2 pp, high confidence
168	K-Exaone	82.5%	estimated ± 2.2 pp, high confidence
169	GLM-4.5-Air	82.4%	estimated ± 2.2 pp, high confidence
170	Solar Pro 2	82.4%	estimated ± 2.2 pp, high confidence
171	GPT-OSS 20B	82.4%	estimated ± 2.2 pp, high confidence
172	Claude Sonnet 4.5 Thinking	82.3%	estimated ± 1.8 pp, high confidence
173	Quasar 438B	82.3%	estimated ± 2.2 pp, medium confidence
174	Llama 4 Scout	82.2%	estimated ± 2.2 pp, medium confidence
175	GLM-4.6	82.2%	measured
176	K-EXAONE 2.0	82.2%	estimated ± 1.6 pp, medium confidence
177	Qwen3-Omni-30B-A3B-Thinking	82.1%	estimated ± 2.2 pp, medium confidence
178	Ling 3.0 Flash VL	82.1%	estimated ± 2.2 pp, medium confidence
179	Qwen3-Omni-30B-A3B-Instruct	82.1%	estimated ± 2.2 pp, medium confidence
180	Phi-4	82.0%	estimated ± 2.2 pp, medium confidence
181	Ling 3.0 Flash	82.0%	measured
182	Gemma 3 27B	81.8%	estimated ± 2.2 pp, medium confidence
183	Sarvam 30B	81.8%	estimated ± 2.2 pp, medium confidence
184	Celeris-1	81.5%	estimated ± 2.2 pp, medium confidence
185	Exaone 4.0 32B	81.4%	estimated ± 2.2 pp, medium confidence
186	GLM-4.5	81.2%	measured
187	LFM2.5-8B-A1B	81.2%	estimated ± 2.2 pp, medium confidence
188	Command A+	81.1%	estimated ± 2.2 pp, medium confidence
189	Ling 3.0 Tiny	81.0%	estimated ± 2.2 pp, medium confidence
190	Gemma 4 31B	80.6%	estimated ± 1.2 pp, high confidence
191	LFM2.5-VL-1.6B-Extract	80.5%	estimated ± 2.2 pp, medium confidence
192	MAI-Thinking-1	80.4%	estimated ± 1.2 pp, high confidence
193	Qwen3.5-35B-A3B	80.4%	estimated ± 1.2 pp, high confidence
194	MiniMax M2.7	80.4%	measured
195	Granite-4.0-H-1B	80.4%	estimated ± 2.2 pp, medium confidence
196	Exaone 4.0 1.2B	80.3%	estimated ± 2.2 pp, medium confidence
197	Mercury 2.5	80.3%	estimated ± 1.6 pp, medium confidence
198	LFM2.5-2.6B	80.2%	estimated ± 2.2 pp, medium confidence
199	Granite-4.0-350M	80.1%	estimated ± 2.2 pp, medium confidence
200	Granite-4.0-H-350M	80.1%	estimated ± 2.2 pp, medium confidence
201	Ling 3.0 Flash FP8	80.0%	estimated ± 1.2 pp, high confidence
202	MiMo-V2-Flash	79.5%	estimated ± 1.2 pp, high confidence
203	Claude Sonnet 4.5	78.8%	estimated ± 1.2 pp, high confidence
204	Claude Haiku 4.5	78.7%	measured
205	Trinity-Large-Thinking	78.6%	estimated ± 1.6 pp, medium confidence
206	Gemini 2.5 Pro	78.0%	estimated ± 1.2 pp, high confidence
207	GPT-5.4 nano	77.2%	measured
208	o1-preview	77.0%	estimated ± 2.6 pp, low confidence
209	Mistral Medium 3.5 128B	75.3%	measured
210	MiniCPM5-2B	74.7%	estimated ± 1.6 pp, medium confidence
211	LongCat-Flash-Lite-Sparse	74.2%	estimated ± 1.6 pp, medium confidence
212	Trinity-Large-Preview	69.9%	estimated ± 1.6 pp, medium confidence
213	Laguna XS.2	69.1%	measured
214	Laguna M.1	68.8%	measured
215	o1-pro	65.7%	estimated ± 1.2 pp, medium confidence
216	Gemma 4 12B	64.9%	estimated ± 1.2 pp, medium confidence
217	Gemma 4 26B A4B	61.2%	estimated ± 1.9 pp, medium confidence
218	Qwen3 235B 2507	59.4%	estimated ± 1.2 pp, medium confidence
219	o3-mini	58.0%	estimated ± 1.2 pp, medium confidence
220	LLaDA2.2-mini	54.8%	estimated ± 1.6 pp, medium confidence
221	o1	50.9%	estimated ± 1.2 pp, medium confidence
222	Nemotron 3.5 Lightning 30B A3B NVFP4	50.2%	estimated ± 1.2 pp, medium confidence
223	MiniCPM5-1B	36.5%	estimated ± 1.6 pp, medium confidence
224	Nemotron 3 Nano Omni 30B A3B	33.7%	estimated ± 1.2 pp, medium confidence
225	ZAYA1-8B	28.2%	estimated ± 1.2 pp, medium confidence
226	Granite 4.2 30B	12.7%	estimated ± 1.2 pp, medium confidence
227	GPT-4.1	12.4%	estimated ± 1.2 pp, medium confidence
228	GPT-4.1 mini	8.2%	estimated ± 1.2 pp, medium confidence
229	Granite 4.2 8B	8.1%	estimated ± 1.2 pp, medium confidence
230	Claude 3.5 Sonnet	3.0%	estimated ± 1.2 pp, medium confidence
231	DeepSeek V3	2.8%	estimated ± 1.2 pp, medium confidence
232	Ling 2.6 Flash	2.7%	estimated ± 1.2 pp, medium confidence
233	Gemma 4 E4B	2.5%	estimated ± 1.2 pp, medium confidence
234	Mellum2-12B-A2.5B-Thinking	2.0%	estimated ± 1.2 pp, medium confidence
235	ZAYA1-74B-Preview	1.9%	estimated ± 1.2 pp, medium confidence
236	Granite 4.2 3B	1.1%	estimated ± 1.2 pp, medium confidence
237	GPT-4.1 nano	0.4%	estimated ± 1.2 pp, medium confidence
238	Gemma 4 E2B	0.1%	estimated ± 1.2 pp, medium confidence
239	Soofi S 30B-A3B	0.1%	estimated ± 1.2 pp, medium confidence
240	Mellum2-12B-A2.5B-Instruct	0.0%	estimated ± 1.2 pp, medium confidence
241	LFM2.5-VL-450M	0.0%	estimated ± 1.2 pp, medium confidence
242	LFM2.5-230M	0.0%	estimated ± 1.2 pp, medium confidence

Vals MMLU-Pro leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General