omniscience Hallucination Rate

113 models evaluated

#	Model	Provider	Type	Score
1	Command A+	Cohere	Open	14.1%
2	Qwen3.7 Max	Alibaba	Closed	22.9%
3	MiMo-V2.5-Pro	Xiaomi	Closed	24.5%
4	Grok 4.3	xAI	Closed	25.0%
5	GLM-5.1	Z.AI	Open	29.4%
6	MiMo-V2-Pro	Xiaomi	Closed	29.9%
7	Gemma 4 E4B	Google	Open	31.3%
8	Qwen3.6 Plus	Alibaba	Closed	32.0%
9	Gemma 4 E2B	Google	Open	32.9%
10	GLM-5	Z.AI	Open	34.0%
11	MiniMax M2.7	MiniMax	Open	34.4%
12	Claude Opus 4.7 (Adaptive)	Anthropic	Closed	36.2%
13	GPT-4o	OpenAI	Closed	37.9%
14	Kimi K2.6	Moonshot AI	Open	39.3%
15	Claude 4 Sonnet	Anthropic	Closed	40.8%
16	Qwen 3.6 Max (preview)	Alibaba	Closed	44.2%
17	MiMo-V2-Omni	Xiaomi	Closed	44.4%
18	Qwen3.6-27B	Alibaba	Open	48.3%
19	Qwen3.6-35B-A3B	Alibaba	Open	49.7%
20	Gemini 3.1 Pro	Google	Closed	49.9%
21	Llama 3.1 405B	Meta	Open	51.0%
22	GPT-5.1	OpenAI	Closed	51.3%
23	Claude Opus 4.7	Anthropic	Closed	51.9%
24	Claude Opus 4.5 Thinking	Anthropic	Closed	59.8%
25	Gemini 3.5 Flash	Google	Closed	60.7%
26	Mistral Medium 3	Mistral	Closed	60.9%
27	Claude Opus 4.6 (Adaptive)	Anthropic	Closed	61.3%
28	GLM-5-Turbo	Z.AI	Closed	62.2%
29	Grok 4	xAI	Closed	64.2%
30	Kimi K2.5 (Reasoning)	Moonshot AI	Closed	64.6%
31	Kimi K2.5	Moonshot AI	Open	64.6%
32	Claude Sonnet 4.6	Anthropic	Closed	65.9%
33	Grok 4 Fast (Reasoning)	xAI	Closed	66.0%
34	GLM-4.6	Z.AI	Open	66.1%
35	Mistral Small 4 (Reasoning)	Mistral	Open	66.8%
36	Mistral Small 4	Mistral	Open	66.8%
37	Mistral Large 2	Mistral	Closed	67.8%
38	GLM-5V-Turbo	Z.AI	Closed	67.9%
39	o1	OpenAI	Closed	69.3%
40	Grok 4.1 Fast (Reasoning)	xAI	Closed	72.4%
41	GPT-5.2-Codex	OpenAI	Closed	72.8%
42	Muse Spark	Meta	Closed	73.2%
43	GPT-5.4 nano	OpenAI	Closed	73.6%
44	Kimi K2	Moonshot AI	Closed	74.2%
45	GPT-5.1-Codex-Max	OpenAI	Closed	74.4%
46	GPT-5.1-Codex	OpenAI	Closed	74.4%
47	MiMo-V2-Flash	Xiaomi	Open	75.1%
48	Claude Opus 4.5	Anthropic	Closed	75.4%
49	Claude Opus 4.6	Anthropic	Closed	76.0%
50	Granite-4.0-350M	IBM	Open	77.8%
51	Nova Pro	Amazon	Closed	77.9%
52	Claude 3 Haiku	Anthropic	Closed	78.2%
53	Llama 4 Scout	Meta	Open	78.3%
54	Grok Code Fast 1	xAI	Closed	78.5%
55	GPT-4.1	OpenAI	Closed	79.6%
56	GPT-5.2	OpenAI	Closed	79.7%
57	Qwen3.5-27B	Alibaba	Open	79.7%
58	Qwen3.5 397B	Alibaba	Open	79.8%
59	GPT-5 (medium)	OpenAI	Closed	80.1%
60	DeepSeek V3.1 (Reasoning)	DeepSeek	Open	80.3%
61	GPT-4.1 nano	OpenAI	Closed	80.4%
62	Phi-4	Microsoft	Open	80.5%
63	Gemma 4 26B A4B	Google	Open	80.9%
64	Exaone 4.0 32B	LG AI Research	Open	81.0%
65	Gemini 3.1 Flash-Lite	Google	Closed	81.6%
66	Gemma 4 31B	Google	Open	81.6%
67	Nemotron Ultra 253B	NVIDIA	Open	81.7%
68	Grok 4.1 Fast	xAI	Closed	81.8%
69	GPT-4.1 mini	OpenAI	Closed	82.0%
70	Mistral Medium 3.5 128B	Mistral	Open	82.0%
71	GPT-5 (high)	OpenAI	Closed	82.1%
72	Nemotron 3 Nano Omni 30B A3B	NVIDIA	Open	83.1%
73	Granite-4.0-H-1B	IBM	Open	83.4%
74	DeepSeek V3.1	DeepSeek	Open	83.5%
75	Mistral Large 3	Mistral	Closed	83.7%
76	Qwen3.5-35B-A3B	Alibaba	Open	84.0%
77	DeepSeek-R1	DeepSeek	Open	84.0%
78	GPT-5.5	OpenAI	Closed	85.5%
79	Qwen3.5-122B-A10B	Alibaba	Open	85.5%
80	Trinity-Large-Preview	Arcee AI	Open	86.6%
81	Trinity-Large-Thinking	Arcee AI	Open	86.6%
82	GPT-5.3 Codex	OpenAI	Closed	86.9%
83	Hy3 Preview	Tencent	Open	86.9%
84	o3	OpenAI	Closed	87.1%
85	Llama 4 Maverick	Meta	Open	87.3%
86	Gemini 2.5 Pro	Google	Closed	87.4%
87	GPT-5.4	OpenAI	Closed	88.6%
88	DeepSeek V4 Pro (High)	DeepSeek	Open	88.6%
89	Qwen3.5 397B (Reasoning)	Alibaba	Open	89.1%
90	K-Exaone	LG AI Research	Closed	89.1%
91	DeepSeek V3	DeepSeek	Open	89.4%
92	Qwen3 Max	Alibaba	Closed	89.4%
93	Gemma 3 27B	Google	Open	89.5%
94	DeepSeek V4 Flash (High)	DeepSeek	Open	89.7%
95	GPT-5.4 mini	OpenAI	Closed	89.8%
96	Gemini 3 Flash	Google	Closed	90.2%
97	GLM-4.7	Z.AI	Open	90.3%
98	Gemini 3 Pro	Google	Closed	90.9%
99	Nemotron 3 Nano 30B	NVIDIA	Open	90.9%
100	GPT-OSS 120B	OpenAI	Open	91.2%
101	Solar Pro 2	Upstage	Closed	91.5%
102	Exaone 4.0 1.2B	LG AI Research	Open	91.5%
103	GLM-4.5-Air	Z.AI	Closed	92.3%
104	Gemini 2.5 Flash	Google	Closed	93.3%
105	DeepSeek V3.2	DeepSeek	Open	93.5%
106	Sarvam 105B	Sarvam	Open	93.5%
107	Granite-4.0-1B	IBM	Open	93.5%
108	DeepSeek V4 Pro (Max)	DeepSeek	Open	94.0%
109	GPT-OSS 20B	OpenAI	Open	94.1%
110	Granite-4.0-H-350M	IBM	Open	94.4%
111	DeepSeek V4 Flash (Max)	DeepSeek	Open	95.8%
112	Ling 2.6 Flash	InclusionAI	Open	95.8%
113	Sarvam 30B	Sarvam	Open	97.0%