AA-Omniscience Hallucination Rate leaderboard

A display-only Artificial Analysis factuality metric for the rate of incorrect answers among non-correct responses.

Models ranked by their published Artificial Analysis Omniscience Hallucination Rate value

Rank	Model	Provider	Hallucination rate
1	Qwen3-Omni-30B-A3B-Instruct	Alibaba	97.6
2	Ling 2.6 Flash	InclusionAI	96.7
3	DeepSeek V4.1 Flash	DeepSeek	96.5
4	SASarvam 30B	Sarvam	96.3
5	LFM2.5-VL-1.6B-Extract	LiquidAI	95.7
6	DeepSeek V4 Pro 0813	DeepSeek	94.1
6	GPT-OSS 20B	OpenAI	94.1
8	Granite-4.0-1B	IBM	93.5
9	SASarvam 105B	Sarvam	93.4
10	DeepSeek V3.2	DeepSeek	93.3
10	GPT-4.1	OpenAI	93.3
12	Gemini 2.5 Flash	Google	93.0
12	GLM-4.7	Z.AI	93.0
12	Solar Pro 2	Upstage	93.0
15	GLM-4.5-Air	Z.AI	92.9
16	CECeleris-1	Celeris	92.8
17	GPT-4.1 mini	OpenAI	92.7
18	GPT-5.6 Luna	OpenAI	92.6
19	Gemini 3 Flash	Google	92.4
20	GPT-5.6 Sol	OpenAI	92.2
21	Gemma 3 27B	Google	92.1
22	DeepSeek V4 Flash 0731	DeepSeek	91.7
22	Exaone 4.0 1.2B	LG AI Research	91.7
22	GPT-5.4	OpenAI	91.7
25	Gemini 3 Pro	Google	91.5
26	Gemini 2.5 Pro	Google	90.9
27	GPT-OSS 120B	OpenAI	90.8
28	GPT-5.4 mini	OpenAI	90.2
28	FAUltravox v0.6 Llama 3.3 70B	Fixie AI	90.2
30	DeepSeek V3	DeepSeek	90.0
31	Qwen3 Max	Alibaba	89.9
32	GPT-5.3 Codex	OpenAI	89.2
33	GPT-5.5	OpenAI	89.0
33	Qwen3-Omni-30B-A3B-Thinking	Alibaba	89.0
35	Llama 4 Maverick	Meta	88.9
36	K-Exaone	LG AI Research	88.8
37	Solar Pro 3	Upstage	88.2
38	Granite-4.0-H-350M	IBM	88.1
38	o3	OpenAI	88.1
40	GPT-5.6 Terra	OpenAI	87.9
41	Qwen3.5-122B-A10B	Alibaba	87.1
42	Nemotron 3 Super 100B	NVIDIA	87.0
43	Gemma 4 26B A4B	Google	86.4
44	Mistral Large 3	Mistral	86.0
45	Trinity-Large-Preview	Arcee AI	85.9
45	Trinity-Large-Thinking	Arcee AI	85.9
47	DeepSeek V3.1	DeepSeek	85.7
47	Nemotron 3 Nano Omni 30B A3B	NVIDIA	85.7
49	Qwen3.5-35B-A3B	Alibaba	85.4
50	Gemma 4 31B	Google	85.0
50	Step 3.7 Flash	StepFun	85.0
52	Muse Spark	Meta	84.2
53	DeepSeek-R1	DeepSeek	83.4
54	Nemotron 3 Nano 30B	NVIDIA	83.3
55	GPT-5 (medium)	OpenAI	83.2
55	North Mini Code	Cohere	83.2
57	Qwen3.5 397B	Alibaba	82.7
57	Qwen3.5 397B (Reasoning)	Alibaba	82.7
59	GPT-4.1 nano	OpenAI	82.6
60	DeepSeek V3.1 (Reasoning)	DeepSeek	82.5
61	Kimi K2.7 Code	Moonshot AI	82.4
62	Grok 4.1 Fast	xAI	82.3
63	GPT-5 (high)	OpenAI	82.2
64	Exaone 4.0 32B	LG AI Research	82.1
65	Muse Glimmer 30B	Meta	81.9
66	Granite-4.0-H-1B	IBM	81.7
67	Mistral Medium 3.5 128B	Mistral	81.6
68	Qwen3.5-27B	Alibaba	81.5
69	GPT-5.2	OpenAI	81.2
69	Nemotron Ultra 253B	NVIDIA	81.2
69	Phi-4	Microsoft	81.2
72	Gemma 4 12B	Google	81.0
73	Claude 3 Haiku	Anthropic	80.5
74	Claude Opus 4.6	Anthropic	80.1
75	Llama 4 Scout	Meta	79.4
76	Grok Code Fast 1	xAI	79.3
77	APApodex 1.1	Apodex	78.4
77	APApodex 1.1 Mini	Apodex	78.4
79	Nova Pro	Amazon	77.7
80	GPT-5.1-Codex	OpenAI	77.2
80	GPT-5.1-Codex-Max	OpenAI	77.2
82	Kimi K2	Moonshot AI	76.6
83	Claude Opus 4.5	Anthropic	76.2
84	Granite-4.0-350M	IBM	76.1
85	MiMo-V2-Flash	Xiaomi	76.0
86	GPT-5.4 nano	OpenAI	74.2
87	Hy3	Tencent	74.1
88	GPT-5.2-Codex	OpenAI	73.4
88	Grok 4.1 Fast (Reasoning)	xAI	73.4
90	Hy3 Preview	Tencent	73.0
91	Claude Fable 5.1	Anthropic	72.6
92	o1	OpenAI	69.6
93	GLM-5V-Turbo	Z.AI	68.8
94	Claude Sonnet 4.6	Anthropic	68.5
95	Grok 4 Fast (Reasoning)	xAI	68.3
96	TMInkling	Thinking Machines Lab	67.7
96	Mistral Large 2	Mistral	67.7
98	GLM-4.6	Z.AI	67.6
99	Mercury 2.5	Inception	67.0
100	Mistral Small 4	Mistral	66.5
100	Mistral Small 4 (Reasoning)	Mistral	66.5
102	Kimi K2.5	Moonshot AI	65.7
102	Kimi K2.5 (Reasoning)	Moonshot AI	65.7
104	Gemini 3.7 Flash	Google	64.5
104	Grok 4	xAI	64.5
106	Claude Fable 5	Anthropic	63.6
107	TMInkling-Small	Thinking Machines Lab	63.0
108	Claude Opus 4.6 (Adaptive)	Anthropic	62.8
109	GLM-5-Turbo	Z.AI	62.6
110	Claude Opus 4.5 Thinking	Anthropic	61.0
111	Mistral Medium 3	Mistral	60.9
112	Claude Opus 5	Anthropic	60.8
113	Gemini 3.5 Flash	Google	60.7
114	Gemini 3.6 Flash	Google	55.6
115	Gemini 3.8 Flash	Google	55.2
116	Claude Opus 4.7	Anthropic	54.1
116	Grok 4.5	xAI	54.1
118	Kimi K3	Moonshot AI	53.2
119	Llama 3.1 405B	Meta	52.4
120	GPT-5.1	OpenAI	51.9
121	GPT-6 Astra	OpenAI	51.3
122	Gemini 3.1 Pro	Google	50.9
123	Qwen3.6-35B-A3B	Alibaba	50.5
124	Muse Spark 1.1	Meta	50.0
125	Qwen3.6-27B	Alibaba	49.3
126	MiMo-V2-Omni	Xiaomi	48.9
127	LFM2.5-8B-A1B	LiquidAI	46.9
128	Qwen 3.6 Max (preview)	Alibaba	46.2
129	Qwen3.8-Flash-Next	Alibaba	45.3
130	Ling 3.0 Flash	InclusionAI	44.1
130	Ling 3.0 Flash FP8	InclusionAI	44.1
132	Claude Opus 4.7 (Adaptive)	Anthropic	42.3
133	Claude 4 Sonnet	Anthropic	41.0
134	Kimi K2.6	Moonshot AI	40.5
135	Claude Sonnet 5	Anthropic	39.4
136	Claude Opus 4.8	Anthropic	39.3
137	GPT-4o	OpenAI	37.9
138	Nemotron 3.5 Lightning 30B A3B NVFP4	NVIDIA	37.6
139	MiniMax M2.7	MiniMax	35.6
140	GLM-5	Z.AI	35.3
141	Qwen3.6 Plus	Alibaba	34.6
142	Gemini 3.5 Flash-Lite	Google	34.4
143	Grok 4.6	xAI	34.3
144	Muse Spark 1.2	Meta	33.3
145	STA.X K2	SK Telecom	33.0
146	Muse Spark 1.3	Meta	32.9
147	Gemma 4 E2B	Google	32.4
148	Granite 4.2 8B	IBM	32.0
149	Gemma 4 E4B	Google	30.9
150	Ling 3.0 Tiny	InclusionAI	30.5
151	Qwen3.8-27B	Alibaba	30.3
152	MiMo-V2-Pro	Xiaomi	30.0
153	GLM-5.1	Z.AI	29.9
154	Nemotron 3 Ultra	NVIDIA	29.7
155	GLM-5.3	Z.AI	29.6
156	Qwen3.8 Max Preview	Alibaba	28.8
157	Qwen3.7 Plus	Alibaba	27.7
158	GLM-5.2	Z.AI	26.3
158	Granite 4.2 3B	IBM	26.3
160	Granite 4.2 30B	IBM	25.6
160	Qwen3.7 Max	Alibaba	25.6
162	Grok 4.3	xAI	25.0
163	MiMo-V2.5-Pro	Xiaomi	24.7
164	Solar Pro 4	Upstage	24.4
165	K-EXAONE 2.0	LG AI Research	22.6
166	Ling 3.0 Flash VL	InclusionAI	22.0
167	OPMiniCPM5-2B	OpenBMB	21.9
168	MCQuasar 438B	Multiverse Computing	21.4
169	MiniMax M3	MiniMax	18.4
170	LFM2.5-2.6B	LiquidAI	16.0
171	Command A+	Cohere	14.2

Evidence key: ObservedLast good

Rows are ordered by the value Artificial Analysis model benchmarks published, highest first. The source does not state whether a higher value is the better result, so this page does not either: for a benchmark that measures a rate of failure, read the table from the bottom.