Knowledge & reasoning

MMLU-Pro (Arcee) leaderboard

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

As of 2026-10-07, the highest measured score on MMLU-Pro (Arcee) is 89.1% by Claude Opus 4.6. 220 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	Claude Fable 5	100.0%	estimated ± 4.0 pp, low confidence
2	Claude Fable 5.1	100.0%	estimated ± 4.0 pp, low confidence
3	Claude Haiku 5.5	100.0%	estimated ± 4.0 pp, low confidence
4	Claude Opus 4.6 (Adaptive)	100.0%	estimated ± 4.0 pp, low confidence
5	Claude Opus 5	100.0%	estimated ± 4.0 pp, low confidence
6	Claude Opus 5.5	100.0%	estimated ± 4.0 pp, low confidence
7	Claude Sonnet 5	100.0%	estimated ± 4.0 pp, low confidence
8	Claude Sonnet 5.5	100.0%	estimated ± 4.0 pp, low confidence
9	Gemini 3.6 Flash	100.0%	estimated ± 4.0 pp, low confidence
10	Gemini 3.7 Flash	100.0%	estimated ± 4.0 pp, low confidence
11	Gemini 3.8 Flash	100.0%	estimated ± 4.0 pp, low confidence
12	Gemini 3 Pro	100.0%	estimated ± 4.0 pp, low confidence
13	Gemini 3 Pro Deep Think	100.0%	estimated ± 4.0 pp, low confidence
14	Gemini 4 Argon	100.0%	estimated ± 4.0 pp, low confidence
15	GLM-5.3	100.0%	estimated ± 4.0 pp, low confidence
16	GLM-5.3-Flash	100.0%	estimated ± 4.0 pp, low confidence
17	GPT-5.2	100.0%	estimated ± 4.0 pp, low confidence
18	GPT-5.2-Codex	100.0%	estimated ± 4.0 pp, low confidence
19	GPT-5.3 Codex	100.0%	estimated ± 4.0 pp, low confidence
20	GPT-5.4 mini	100.0%	estimated ± 4.0 pp, low confidence
21	GPT-5.4 nano	100.0%	estimated ± 4.0 pp, low confidence
22	GPT-5.4 Pro	100.0%	estimated ± 4.0 pp, low confidence
23	GPT-5.5 Pro	100.0%	estimated ± 4.0 pp, low confidence
24	GPT-6.1 Sol	100.0%	estimated ± 4.0 pp, low confidence
25	GPT-6 Luna	100.0%	estimated ± 4.0 pp, low confidence
26	GPT-6 Sol	100.0%	estimated ± 4.0 pp, low confidence
27	Grok 4.3	100.0%	estimated ± 4.0 pp, low confidence
28	Grok 4.5	100.0%	estimated ± 4.0 pp, low confidence
29	Grok 4.6	100.0%	estimated ± 4.0 pp, low confidence
30	Grok 4.7	100.0%	estimated ± 4.0 pp, low confidence
31	Kimi K2.7 Code	100.0%	estimated ± 4.0 pp, low confidence
32	Ling 3.1 Flash	100.0%	estimated ± 4.0 pp, low confidence
33	MiMo-V2.6-Flash	100.0%	estimated ± 4.0 pp, low confidence
34	MiMo-V2.6-Pro	100.0%	estimated ± 4.0 pp, low confidence
35	Mistral Large 4	100.0%	estimated ± 4.0 pp, low confidence
36	Muse Spark 1.1	100.0%	estimated ± 4.0 pp, low confidence
37	Muse Spark 1.2	100.0%	estimated ± 4.0 pp, low confidence
38	Muse Spark 1.3	100.0%	estimated ± 4.0 pp, low confidence
39	Quasar 438B	100.0%	estimated ± 4.0 pp, low confidence
40	Qwen3.8 Max Preview	100.0%	estimated ± 4.0 pp, low confidence
41	GPT-5.1-Codex	98.5%	estimated ± 4.0 pp, low confidence
42	GPT-5.1-Codex-Max	98.5%	estimated ± 4.0 pp, low confidence
43	GPT-5 (high)	98.5%	estimated ± 4.0 pp, low confidence
44	Claude Opus 4.7	96.3%	estimated ± 4.0 pp, low confidence
45	GPT-5.1	95.5%	estimated ± 4.0 pp, low confidence
46	Hy3	95.5%	estimated ± 4.0 pp, low confidence
47	Apodex 1.1	94.4%	estimated ± 4.0 pp, low confidence
48	Apodex 1.1 Mini	94.4%	estimated ± 4.0 pp, low confidence
49	Claude Opus 4.5 Thinking	94.4%	estimated ± 4.0 pp, low confidence
50	MiMo-V2.5-Pro	92.1%	estimated ± 4.0 pp, low confidence
51	MiniMax M3	91.0%	estimated ± 4.0 pp, low confidence
52	Qwen 3.6 Max (preview)	91.0%	estimated ± 4.0 pp, low confidence
53	Claude Opus 4.6	89.1%	measured
54	GPT-6 Astra	88.8%	estimated ± 3.4 pp, low confidence
55	Kimi K2.5 (Reasoning)	88.8%	estimated ± 4.0 pp, medium confidence
56	Nemotron 3 Super 100B	88.8%	estimated ± 4.0 pp, medium confidence
57	Sakana Fugu	88.7%	estimated ± 3.4 pp, low confidence
58	Sakana Fugu-Ultra	88.7%	estimated ± 3.4 pp, low confidence
59	GPT-5.6 Sol	88.4%	estimated ± 3.4 pp, low confidence
60	Gemini 3.1 Pro	88.3%	estimated ± 3.4 pp, low confidence
61	Claude Opus 4.7 (Adaptive)	88.2%	estimated ± 3.4 pp, low confidence
62	Claude Opus 4.8	88.1%	estimated ± 3.4 pp, low confidence
63	GPT-5.5	88.1%	estimated ± 3.4 pp, low confidence
64	Kimi K3	88.0%	estimated ± 3.4 pp, low confidence
65	Step 5 Preview	88.0%	estimated ± 3.4 pp, low confidence
66	Grok 4.1 Fast (Reasoning)	88.0%	estimated ± 4.0 pp, medium confidence
67	Grok 4 Fast (Reasoning)	88.0%	estimated ± 4.0 pp, medium confidence
68	Qwen3.6 Plus	88.0%	estimated ± 4.0 pp, medium confidence
69	GPT-5.6 Terra	87.8%	estimated ± 3.4 pp, low confidence
70	GPT-5.4	87.8%	estimated ± 3.4 pp, low confidence
71	Ornith-1.5-397B	87.8%	estimated ± 3.4 pp, low confidence
72	Gemini 3.5 Flash	87.8%	estimated ± 3.4 pp, low confidence
73	Qwen3.8 Max	87.7%	estimated ± 3.4 pp, low confidence
74	Pareto 26.10 Preview	87.7%	estimated ± 3.4 pp, low confidence
75	Qwen3.7 Max	87.7%	estimated ± 3.4 pp, low confidence
76	GPT-5.6 Luna	87.6%	estimated ± 3.4 pp, low confidence
77	Hy4 preview	87.6%	estimated ± 3.4 pp, low confidence
78	Qwen3.8-Flash-Next	87.4%	estimated ± 3.4 pp, low confidence
79	GLM-5.2	87.3%	estimated ± 3.4 pp, low confidence
80	Qwen3.8-Omni-Flash	87.2%	estimated ± 3.4 pp, low confidence
81	DeepSeek V4.1 Flash	87.2%	estimated ± 3.4 pp, low confidence
82	Kimi K2.5	87.1%	measured
83	Kimi K2.6	87.0%	estimated ± 3.4 pp, low confidence
84	Beam	87.0%	estimated ± 3.4 pp, low confidence
85	Qwen3.7 Plus	87.0%	estimated ± 3.4 pp, low confidence
86	DeepSeek V4 Pro 0813	86.9%	estimated ± 3.4 pp, low confidence
87	Gemini 2.5 Pro	86.9%	estimated ± 4.0 pp, medium confidence
88	Muse Glimmer 30B	86.9%	estimated ± 4.0 pp, medium confidence
89	Interfaze Beta	86.8%	estimated ± 3.4 pp, low confidence
90	Inkling-Small	86.7%	estimated ± 3.4 pp, low confidence
91	Muse Spark	86.7%	estimated ± 3.4 pp, low confidence
92	Ornith-1.5-35B-A3B	86.6%	estimated ± 3.4 pp, medium confidence
93	Qwen3.8-27B	86.6%	estimated ± 3.4 pp, medium confidence
94	Solar Pro 4	86.5%	estimated ± 3.4 pp, medium confidence
95	Grok 4.20	86.4%	estimated ± 3.4 pp, medium confidence
96	DeepSeek V4 Flash 0731	86.2%	estimated ± 3.4 pp, medium confidence
97	Inkling	86.2%	estimated ± 3.4 pp, medium confidence
98	Hy3 Preview	85.9%	estimated ± 3.4 pp, medium confidence
99	Nemotron 3 Ultra	85.8%	estimated ± 3.4 pp, medium confidence
100	GLM-5	85.8%	measured
101	Step 3.7 Flash	85.8%	estimated ± 4.0 pp, medium confidence
102	Ornith-1.5-9B	85.6%	estimated ± 3.4 pp, medium confidence
103	Solar Open 2	85.6%	estimated ± 3.4 pp, medium confidence
104	GLM-5.1	85.6%	estimated ± 3.4 pp, medium confidence
105	Ternary Bonsai 2 27B	85.4%	estimated ± 3.4 pp, medium confidence
106	A.X K2	85.3%	estimated ± 3.4 pp, medium confidence
107	Ling 3.0 Flash	85.1%	estimated ± 3.4 pp, medium confidence
108	MAI-Thinking-1	84.8%	estimated ± 3.4 pp, medium confidence
109	Ling 3.0 Flash FP8	84.8%	estimated ± 3.4 pp, medium confidence
110	DeepSeek V3.1 (Reasoning)	84.6%	estimated ± 4.0 pp, medium confidence
111	Grok 4	84.6%	estimated ± 4.0 pp, medium confidence
112	Ling 3.0 Flash VL	84.6%	estimated ± 4.0 pp, medium confidence
113	o3-pro	84.3%	estimated ± 5.4 pp, low confidence
114	K-EXAONE 2.0	84.1%	estimated ± 3.4 pp, medium confidence
115	GLM-4.7	83.5%	estimated ± 4.0 pp, medium confidence
116	Trinity-Large-Thinking	83.4%	measured
117	Mercury 2.5	82.8%	estimated ± 3.4 pp, medium confidence
118	Gemma 4 12B	82.8%	estimated ± 3.4 pp, medium confidence
119	Claude 4.1 Opus	82.7%	estimated ± 5.4 pp, low confidence
120	DeepSeek-R1	82.4%	estimated ± 4.0 pp, medium confidence
121	Gemini 2.5 Flash	82.4%	estimated ± 4.0 pp, medium confidence
122	Gemini 3 Flash	82.4%	estimated ± 4.0 pp, medium confidence
123	Gemma 4 31B	82.4%	estimated ± 4.0 pp, medium confidence
124	GPT-OSS 20B	82.4%	estimated ± 4.0 pp, medium confidence
125	Nemotron 3.5 Lightning 30B A3B NVFP4	81.4%	estimated ± 3.4 pp, medium confidence
126	Claude 4 Sonnet	81.3%	estimated ± 4.0 pp, medium confidence
127	GPT-OSS 120B	81.3%	estimated ± 4.0 pp, medium confidence
128	K-Exaone	81.3%	estimated ± 4.0 pp, medium confidence
129	MiMo-V2-Omni	81.3%	estimated ± 4.0 pp, medium confidence
130	o3	81.3%	estimated ± 4.0 pp, medium confidence
131	Qwen3.6-27B	81.3%	estimated ± 4.0 pp, medium confidence
132	MiniMax M2.7	80.8%	measured
133	Claude Sonnet 4.6	80.5%	estimated ± 4.0 pp, medium confidence
134	DeepSeek V3.2	80.5%	estimated ± 4.0 pp, medium confidence
135	Nemotron 3 Nano 30B	80.5%	estimated ± 4.0 pp, medium confidence
136	Qwen3.5-27B	80.5%	estimated ± 4.0 pp, medium confidence
137	Qwen3.5-35B-A3B	80.5%	estimated ± 4.0 pp, medium confidence
138	Qwen3.5 397B	80.5%	estimated ± 4.0 pp, medium confidence
139	Qwen3.5 397B (Reasoning)	80.5%	estimated ± 4.0 pp, medium confidence
140	Nemotron 3 Nano Omni 30B A3B	80.0%	estimated ± 3.4 pp, medium confidence
141	o3-mini	79.9%	estimated ± 5.4 pp, low confidence
142	o1-pro	79.9%	estimated ± 5.4 pp, low confidence
143	ZAYA1-8B	79.4%	estimated ± 3.4 pp, medium confidence
144	Gemma 4 E4B	79.4%	estimated ± 4.0 pp, medium confidence
145	GLM-5V-Turbo	79.4%	estimated ± 4.0 pp, medium confidence
146	Qwen3.5-122B-A10B	79.4%	estimated ± 4.0 pp, medium confidence
147	o1-preview	79.4%	estimated ± 5.4 pp, low confidence
148	MiniCPM5-2B	79.1%	estimated ± 3.4 pp, medium confidence
149	LongCat-Flash-Lite-Sparse	78.7%	estimated ± 3.4 pp, medium confidence
150	Claude Opus 4.5	78.3%	estimated ± 4.0 pp, low confidence
151	Command A+	78.3%	estimated ± 4.0 pp, low confidence
152	GLM-5-Turbo	78.3%	estimated ± 4.0 pp, low confidence
153	Granite 4.2 30B	78.3%	estimated ± 4.0 pp, low confidence
154	Granite 4.2 8B	78.3%	estimated ± 4.0 pp, low confidence
155	MiMo-V2-Pro	78.3%	estimated ± 4.0 pp, low confidence
156	Mistral Small 4	78.3%	estimated ± 4.0 pp, low confidence
157	Mistral Small 4 (Reasoning)	78.3%	estimated ± 4.0 pp, low confidence
158	North Mini Code	78.3%	estimated ± 4.0 pp, low confidence
159	o1	78.3%	estimated ± 4.0 pp, low confidence
160	Qwen3.6-35B-A3B	78.3%	estimated ± 4.0 pp, low confidence
161	Sarvam 30B	78.3%	estimated ± 4.0 pp, low confidence
162	Claude 3 Opus	78.1%	estimated ± 5.4 pp, low confidence
163	DeepSeek R1 Distill Qwen 32B	78.0%	estimated ± 5.4 pp, low confidence
164	Gemini 1.5 Pro	77.7%	estimated ± 5.4 pp, low confidence
165	GPT-4 Turbo	77.3%	estimated ± 5.4 pp, low confidence
166	Qwen2.5 Coder 32B Instruct	77.2%	estimated ± 5.4 pp, low confidence
167	GPT-4o mini	77.2%	estimated ± 5.4 pp, low confidence
168	Celeris-1	77.1%	estimated ± 4.0 pp, low confidence
169	Claude 3 Haiku	77.1%	estimated ± 4.0 pp, low confidence
170	Claude 4.1 Opus Thinking	77.1%	estimated ± 4.0 pp, low confidence
171	DeepSeek V3	77.1%	estimated ± 4.0 pp, low confidence
172	DeepSeek V3 0324	77.1%	estimated ± 4.0 pp, low confidence
173	DeepSeek V3.1	77.1%	estimated ± 4.0 pp, low confidence
174	Exaone 4.0 1.2B	77.1%	estimated ± 4.0 pp, low confidence
175	Exaone 4.0 32B	77.1%	estimated ± 4.0 pp, low confidence
176	Gemini 3.5 Flash-Lite	77.1%	estimated ± 4.0 pp, low confidence
177	Gemma 3 27B	77.1%	estimated ± 4.0 pp, low confidence
178	Gemma 4 26B A4B	77.1%	estimated ± 4.0 pp, low confidence
179	Gemma 4 E2B	77.1%	estimated ± 4.0 pp, low confidence
180	GLM-4.5-Air	77.1%	estimated ± 4.0 pp, low confidence
181	GLM-4.6	77.1%	estimated ± 4.0 pp, low confidence
182	GPT-4.1	77.1%	estimated ± 4.0 pp, low confidence
183	GPT-4.1 mini	77.1%	estimated ± 4.0 pp, low confidence
184	GPT-4.1 nano	77.1%	estimated ± 4.0 pp, low confidence
185	GPT-4o	77.1%	estimated ± 4.0 pp, low confidence
186	GPT-5 (medium)	77.1%	estimated ± 4.0 pp, low confidence
187	Granite-4.0-350M	77.1%	estimated ± 4.0 pp, low confidence
188	Granite-4.0-H-1B	77.1%	estimated ± 4.0 pp, low confidence
189	Granite-4.0-H-350M	77.1%	estimated ± 4.0 pp, low confidence
190	Granite 4.2 3B	77.1%	estimated ± 4.0 pp, low confidence
191	Grok 4.1 Fast	77.1%	estimated ± 4.0 pp, low confidence
192	Grok Code Fast 1	77.1%	estimated ± 4.0 pp, low confidence
193	Kimi K2	77.1%	estimated ± 4.0 pp, low confidence
194	LFM2.5-2.6B	77.1%	estimated ± 4.0 pp, low confidence
195	LFM2.5-8B-A1B	77.1%	estimated ± 4.0 pp, low confidence
196	LFM2.5-VL-1.6B-Extract	77.1%	estimated ± 4.0 pp, low confidence
197	Ling 2.6 Flash	77.1%	estimated ± 4.0 pp, low confidence
198	Ling 3.0 Tiny	77.1%	estimated ± 4.0 pp, low confidence
199	Llama 3.1 405B	77.1%	estimated ± 4.0 pp, low confidence
200	Llama 4 Maverick	77.1%	estimated ± 4.0 pp, low confidence
201	Llama 4 Scout	77.1%	estimated ± 4.0 pp, low confidence
202	MiMo-V2-Flash	77.1%	estimated ± 4.0 pp, low confidence
203	Mistral Large 2	77.1%	estimated ± 4.0 pp, low confidence
204	Mistral Large 3	77.1%	estimated ± 4.0 pp, low confidence
205	Mistral Medium 3	77.1%	estimated ± 4.0 pp, low confidence
206	Mistral Medium 3.5 128B	77.1%	estimated ± 4.0 pp, low confidence
207	Nemotron Ultra 253B	77.1%	estimated ± 4.0 pp, low confidence
208	Nova Pro	77.1%	estimated ± 4.0 pp, low confidence
209	Phi-4	77.1%	estimated ± 4.0 pp, low confidence
210	Qwen3 Max	77.1%	estimated ± 4.0 pp, low confidence
211	Qwen3-Omni-30B-A3B-Instruct	77.1%	estimated ± 4.0 pp, low confidence
212	Qwen3-Omni-30B-A3B-Thinking	77.1%	estimated ± 4.0 pp, low confidence
213	Sarvam 105B	77.1%	estimated ± 4.0 pp, low confidence
214	Solar Pro 2	77.1%	estimated ± 4.0 pp, low confidence
215	Solar Pro 3	77.1%	estimated ± 4.0 pp, low confidence
216	Ultravox v0.6 Llama 3.3 70B	77.1%	estimated ± 4.0 pp, low confidence
217	Phi-4 Multimodal Instruct	76.8%	estimated ± 5.4 pp, low confidence
218	Gemini 1.0 Pro	76.5%	estimated ± 5.4 pp, low confidence
219	Trinity-Large-Preview	75.2%	measured
220	Mellum2-12B-A2.5B-Thinking	72.6%	estimated ± 3.4 pp, low confidence
221	ZAYA1-74B-Preview	72.4%	estimated ± 3.4 pp, low confidence
222	LLaDA2.2-mini	63.9%	estimated ± 3.4 pp, low confidence
223	Soofi S 30B-A3B	63.1%	estimated ± 3.4 pp, low confidence
224	Mellum2-12B-A2.5B-Instruct	61.2%	estimated ± 3.4 pp, low confidence
225	MiniCPM5-1B	47.0%	estimated ± 3.4 pp, low confidence
226	LFM2.5-230M	46.0%	estimated ± 3.4 pp, low confidence

MMLU-Pro (Arcee) leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General