Knowledge & reasoning

AA-Omniscience Accuracy leaderboard

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

As of 2026-10-07, the highest measured score on AA-Omniscience Accuracy is 67.2% by Claude Fable 5.1. 61 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	Claude Fable 5.1	67.2%	measured
2	Claude Opus 5.5	66.2%	measured
3	Claude Fable 5	65.4%	measured
4	GPT-6 Astra	62.6%	measured
5	GPT-6.1 Sol	62.1%	measured
6	Claude Opus 5	60.9%	measured
7	GPT-5.6 Sol	59.4%	measured
8	Sakana Fugu	58.1%	estimated ± 8.1 pp, medium confidence
9	Sakana Fugu-Ultra	58.1%	estimated ± 8.1 pp, medium confidence
10	GPT-5.5	58.0%	measured
11	GPT-5.5 Pro	56.2%	estimated ± 6.2 pp, medium confidence
12	GPT-5.4 Pro	55.9%	estimated ± 6.2 pp, medium confidence
13	Gemini 3 Pro	55.8%	measured
14	Gemini 3.7 Flash	55.3%	measured
15	dots3-note Preview	55.2%	estimated ± 6.2 pp, medium confidence
16	Gemini 3.1 Pro	54.9%	measured
17	Gemini 3.8 Flash	54.6%	measured
18	GPT-6 Sol	54.5%	measured
19	Claude Sonnet 5.5	54.0%	measured
20	GPT-5.3 Codex	52.9%	measured
21	Muse Spark 1.1	52.1%	measured
22	Gemini 3.5 Flash	51.9%	measured
23	Grok 4.5	51.6%	measured
24	Claude Mythos 5	51.2%	estimated ± 8.1 pp, medium confidence
25	GPT-5.4	50.8%	measured
26	Pareto 26.9	50.3%	estimated ± 8.6 pp, medium confidence
27	Gemini 3.6 Flash	50.0%	measured
28	Gemini 4 Argon	49.9%	measured
29	Muse Spark	49.6%	measured
30	DeepSeek V4 Pro 0813	49.1%	measured
31	Claude Opus 4.7 (Adaptive)	48.9%	measured
32	Claude Opus 4.8	48.8%	measured
33	Grok 4.6	48.2%	measured
34	Grok 4.20	47.6%	estimated ± 4.6 pp, medium confidence
35	Kimi K3	47.6%	measured
36	Grok 4.7	47.4%	measured
37	Claude Opus 4.6 (Adaptive)	47.0%	measured
38	GPT-5.6 Terra	46.8%	measured
39	Claude Opus 4.5 Thinking	46.6%	measured
40	DeepSeek V4.1 Flash	46.4%	measured
41	Claude Haiku 5.5	46.1%	estimated ± 6.9 pp, medium confidence
42	Qwen3.8 Max	46.0%	estimated ± 7.7 pp, medium confidence
43	Claude Opus 4.6	45.8%	measured
44	Gemini 3 Flash	45.8%	measured
45	Muse Spark 1.2	45.4%	measured
46	Ornith-1.5-397B	45.4%	estimated ± 8.1 pp, medium confidence
47	Hy4 preview	45.0%	estimated ± 7.9 pp, medium confidence
48	Claude Opus 4.7	44.7%	measured
49	GPT-5.2	44.3%	measured
50	GPT-6 Luna	43.8%	measured
51	Muse Spark 1.3	43.6%	measured
52	Gemini 3 Pro Deep Think	43.4%	estimated ± 6.2 pp, medium confidence
53	Pareto 26.10 Preview	42.9%	estimated ± 8.6 pp, medium confidence
54	GPT-5.6 Luna	42.7%	measured
55	Inkling	41.6%	measured
56	Step 5 Preview	41.5%	measured
57	GLM-5.3-Flash	41.3%	estimated ± 6.9 pp, medium confidence
58	GPT-5.2-Codex	41.1%	measured
59	Claude Opus 4.5	40.9%	measured
60	Grok 4	40.5%	measured
61	DeepSeek V4 Flash 0731	40.4%	measured
62	GPT-5 (high)	40.3%	measured
63	Claude Sonnet 5	40.1%	measured
64	GPT-5.1-Codex	39.9%	measured
65	GPT-5.1-Codex-Max	39.9%	measured
66	Agents-A1	39.9%	estimated ± 10.5 pp, low confidence
67	Kimi K2.7 Code	39.6%	measured
68	GPT-5 (medium)	39.5%	measured
69	Gemini 2.5 Pro	39.1%	measured
70	Claude Sonnet 4.6	38.6%	measured
71	o3	38.6%	measured
72	Qwen3.8-Omni-Flash	38.3%	estimated ± 8.1 pp, medium confidence
73	Beam	38.2%	estimated ± 8.6 pp, medium confidence
74	Qwen 3.6 Max (preview)	37.9%	measured
75	GPT-5.1	37.7%	measured
76	GPT-5.4 mini	37.5%	measured
77	Kimi K2.5	35.2%	measured
78	Kimi K2.5 (Reasoning)	35.2%	measured
79	MiMo-V2.6-Pro	34.8%	measured
80	Interfaze Beta	34.7%	estimated ± 8.1 pp, medium confidence
81	Grok 4.3	34.6%	measured
82	o1	34.5%	measured
83	GLM-5.3	33.9%	measured
84	Gemini 3.1 Flash-Lite	33.7%	estimated ± 7.7 pp, medium confidence
85	Inkling-Small	33.2%	measured
86	Claude Sonnet 4.5 Thinking	33.0%	estimated ± 6.2 pp, medium confidence
87	Claude Sonnet 4.5	33.0%	estimated ± 6.2 pp, medium confidence
88	Kimi K2.6	32.6%	measured
89	Ornith-1.5-35B-A3B	32.6%	estimated ± 8.1 pp, medium confidence
90	Hy3	32.0%	measured
91	Apodex 1.1	31.7%	measured
92	Apodex 1.1 Mini	31.7%	measured
93	Qwen3.8 Max Preview	31.7%	measured
94	Hy3 Preview	31.5%	measured
95	Solar Open 2	31.1%	estimated ± 8.6 pp, medium confidence
96	Qwen3.7 Max	31.1%	measured
97	DeepSeek-R1	30.5%	measured
98	Gemini 3.5 Flash-Lite	29.5%	measured
99	GLM-4.7	29.3%	measured
100	GLM-5V-Turbo	29.3%	measured
101	Ling 3.1 Flash	29.1%	measured
102	DeepSeek V3.1 (Reasoning)	29.0%	measured
103	GLM-5-Turbo	28.4%	measured
104	o3-pro	27.9%	estimated ± 7.5 pp, medium confidence
105	Seed 2.1 Pro	27.9%	estimated ± 6.0 pp, medium confidence
106	GPT-4.1	27.8%	measured
107	Kimi K2	27.4%	measured
108	Qwen3.5 Flash	27.0%	estimated ± 7.7 pp, medium confidence
109	MiMo-V2.6-Flash	27.0%	measured
110	Muse Glimmer 30B	27.0%	measured
111	MiniMax M2.7	26.8%	measured
112	MiMo-V2-Pro	26.6%	measured
113	Qwen3.6 Plus	26.4%	measured
114	GLM-5	26.3%	measured
115	Ornith-1.5-9B	26.1%	estimated ± 8.1 pp, medium confidence
116	Gemini 2.5 Flash	26.1%	measured
117	Mistral Large 4	25.8%	measured
118	Step 3.7 Flash	25.8%	measured
119	GPT-5.4 nano	25.7%	measured
120	Seed 2.1 Turbo	25.6%	estimated ± 6.0 pp, medium confidence
121	DeepSeek V3	25.5%	measured
122	Grok 4.1 Fast (Reasoning)	25.1%	measured
123	Mistral Large 3	25.0%	measured
124	MiMo-V2.5	24.9%	estimated ± 7.7 pp, medium confidence
125	Llama 4 Maverick	24.9%	measured
126	Claude 4.1 Opus	24.8%	estimated ± 7.5 pp, medium confidence
127	Mistral Medium 3.5 128B	24.7%	measured
128	Qwen3.5 397B	24.5%	measured
129	Qwen3.5 397B (Reasoning)	24.5%	measured
130	Qwen3.8-Flash-Next	24.5%	measured
131	Qwen3.5-122B-A10B	24.4%	measured
132	Qwen3 Max	24.4%	measured
133	DeepSeek V3 0324	24.3%	measured
134	GLM-5.2	24.3%	measured
135	Nemotron 3 Super 100B	24.3%	measured
136	DeepSeek V3.2	24.0%	measured
137	GLM-5.1	23.7%	measured
138	Grok Code Fast 1	23.5%	measured
139	GLM-4.5	23.2%	estimated ± 7.7 pp, medium confidence
140	Llama 3.1 405B	23.2%	measured
141	DeepSeek V3.1	23.1%	measured
142	Grok 4 Fast (Reasoning)	22.8%	measured
143	MAI-Thinking-1	22.8%	estimated ± 8.1 pp, medium confidence
144	Claude 4 Sonnet	22.7%	measured
145	Qwen3.7 Plus	22.5%	measured
146	Trinity-Large-Preview	22.5%	measured
147	Trinity-Large-Thinking	22.5%	measured
148	Qwen3 235B 2507	22.4%	estimated ± 6.0 pp, medium confidence
149	MiMo-V2.5-Pro	22.4%	measured
150	Claude Haiku 4.5	22.2%	estimated ± 7.7 pp, medium confidence
151	Mercury 2.5	22.0%	measured
152	GPT-OSS 120B	21.8%	measured
153	Mistral Small 4	21.7%	measured
154	Mistral Small 4 (Reasoning)	21.7%	measured
155	Laguna XS.2	21.7%	estimated ± 7.7 pp, low confidence
156	Laguna M.1	21.7%	estimated ± 7.7 pp, low confidence
157	Nemotron 3 Ultra	21.6%	measured
158	GLM-4.6	21.4%	measured
159	o3-mini	20.8%	estimated ± 1.8 pp, medium confidence
160	Qwen3.5-27B	20.7%	measured
161	GPT-4.1 mini	20.3%	measured
162	Nemotron Ultra 253B	20.1%	measured
163	Qwen3.5-35B-A3B	20.1%	measured
164	Gemma 4 31B	20.0%	measured
165	GPT-4o	19.9%	measured
166	Mistral Large 2	19.9%	measured
167	Qwen3.6-27B	19.6%	measured
168	o1-pro	19.5%	estimated ± 7.5 pp, medium confidence
169	MiMo-V2-Omni	19.3%	measured
170	Claude 4.1 Opus Thinking	19.3%	estimated ± 6.9 pp, medium confidence
171	Gemma 4 26B A4B	19.1%	measured
172	Ultravox v0.6 Llama 3.3 70B	19.0%	measured
173	North Mini Code	18.9%	measured
174	Solar Pro 4	18.9%	measured
175	Qwen3.6-35B-A3B	18.8%	measured
176	o1-preview	18.7%	estimated ± 7.5 pp, medium confidence
177	A.X K2	18.6%	measured
178	Solar Pro 3	18.5%	measured
179	LongCat-Flash-Lite-Sparse	18.4%	estimated ± 1.8 pp, medium confidence
180	Mistral Medium 3	18.3%	measured
181	Ling 3.0 Flash	18.2%	measured
182	Ling 3.0 Flash FP8	18.2%	measured
183	Claude 3 Haiku	17.6%	measured
184	Sarvam 105B	17.6%	measured
185	Phi-4 Multimodal Instruct	17.5%	estimated ± 6.9 pp, medium confidence
186	DeepSeek R1 Distill Qwen 32B	17.4%	estimated ± 6.9 pp, medium confidence
187	Gemini 1.5 Pro	17.4%	estimated ± 6.9 pp, medium confidence
188	Gemini 1.0 Pro	17.4%	estimated ± 6.9 pp, medium confidence
189	GPT-4o mini	17.4%	estimated ± 6.9 pp, medium confidence
190	Qwen2.5 Coder 32B Instruct	17.3%	estimated ± 6.9 pp, medium confidence
191	GPT-4 Turbo	17.3%	estimated ± 6.9 pp, medium confidence
192	Nemotron 3 Nano 30B	17.3%	measured
193	Claude 3 Opus	17.3%	estimated ± 6.9 pp, medium confidence
194	Grok 4.1 Fast	17.2%	measured
195	Nova Pro	16.9%	measured
196	MiniMax M3	16.7%	measured
197	ZAYA1-8B	16.7%	estimated ± 8.1 pp, medium confidence
198	K-Exaone	16.4%	measured
199	GLM-4.5-Air	16.3%	measured
200	Claude 3.5 Sonnet	16.3%	estimated ± 8.1 pp, medium confidence
201	ZAYA1-74B-Preview	16.3%	estimated ± 8.1 pp, medium confidence
202	Soofi S 30B-A3B	16.2%	estimated ± 8.1 pp, medium confidence
203	LFM2.5-230M	16.2%	estimated ± 8.1 pp, low confidence
204	LFM2.5-VL-450M	16.2%	estimated ± 8.1 pp, low confidence
205	Solar Pro 2	16.1%	measured
206	GPT-OSS 20B	16.0%	measured
207	Gemma 4 12B	15.6%	measured
208	Ling 2.6 Flash	15.6%	measured
209	MiMo-V2-Flash	15.6%	measured
210	Qwen3.8-27B	15.6%	measured
211	Quasar 438B	15.5%	measured
212	Llama 4 Scout	15.2%	measured
213	Nemotron 3 Nano Omni 30B A3B	15.2%	measured
214	Qwen3-Omni-30B-A3B-Thinking	14.6%	measured
215	Ling 3.0 Flash VL	14.4%	measured
216	Nemotron 3.5 Lightning 30B A3B NVFP4	14.4%	measured
217	Qwen3-Omni-30B-A3B-Instruct	14.3%	measured
218	Phi-4	14.1%	measured
219	GPT-4.1 nano	13.7%	measured
220	K-EXAONE 2.0	13.1%	measured
221	Gemma 3 27B	13.0%	measured
222	Sarvam 30B	12.6%	measured
223	Ternary Bonsai 2 27B	11.7%	estimated ± 4.5 pp, medium confidence
224	Granite 4.2 8B	11.2%	measured
225	Celeris-1	11.0%	measured
226	Exaone 4.0 32B	10.6%	measured
227	Granite 4.2 30B	10.1%	measured
228	LFM2.5-8B-A1B	9.4%	measured
229	Granite 4.2 3B	9.2%	measured
230	Mellum2-12B-A2.5B-Thinking	8.9%	estimated ± 4.5 pp, medium confidence
231	Command A+	8.9%	measured
232	Gemma 4 E4B	8.6%	measured
233	Ling 3.0 Tiny	8.5%	measured
234	MiniCPM5-2B	8.4%	measured
235	Gemma 4 E2B	6.6%	measured
236	LFM2.5-VL-1.6B-Extract	5.8%	measured
237	Granite-4.0-H-1B	5.2%	measured
238	Mellum2-12B-A2.5B-Instruct	5.1%	estimated ± 4.5 pp, low confidence
239	Exaone 4.0 1.2B	5.0%	measured
240	LFM2.5-2.6B	4.4%	measured
241	LLaDA2.2-mini	4.1%	estimated ± 8.6 pp, low confidence
242	Granite-4.0-350M	3.9%	measured
243	Granite-4.0-H-350M	3.8%	measured
244	MiniCPM5-1B	3.3%	estimated ± 4.5 pp, low confidence

AA-Omniscience Accuracy leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General