AA-GPQA Diamond leaderboard

A display-only Artificial Analysis GPQA Diamond score.

Models ranked by their published Artificial Analysis GPQA Diamond value

Rank	Model	Provider	Accuracy
1	GPT-6 Astra	OpenAI	96.1
2	Gemini 3.8 Flash	Google	95.3
3	Grok 4.6	xAI	94.9
4	Gemini 3.7 Flash	Google	94.5
5	Gemini 3.1 Pro	Google	94.1
5	GPT-5.6 Sol	OpenAI	94.1
7	Claude Fable 5.1	Anthropic	93.7
8	GPT-5.5	OpenAI	93.5
8	Kimi K3	Moonshot AI	93.5
8	Muse Spark 1.3	Meta	93.5
11	Claude Opus 5	Anthropic	93.2
12	Grok 4.5	xAI	93.1
13	MiniMax M3	MiniMax	92.9
14	DeepSeek V4 Pro 0813	DeepSeek	92.8
14	Gemini 3.6 Flash	Google	92.8
14	Qwen3.8 Max Preview	Alibaba	92.8
17	Claude Fable 5	Anthropic	92.6
18	GPT-5.6 Terra	OpenAI	92.5
19	Qwen3.7 Max	Alibaba	92.3
19	Qwen3.8-Flash-Next	Alibaba	92.3
21	Gemini 3.5 Flash	Google	92.2
22	Claude Opus 4.8	Anthropic	92.0
22	GPT-5.4	OpenAI	92.0
24	GLM-5.3	Z.AI	91.7
25	GPT-5.3 Codex	OpenAI	91.5
26	Claude Opus 4.7 (Adaptive)	Anthropic	91.4
27	GLM-5.3-Flash	Z.AI	91.2
28	Claude Sonnet 5	Anthropic	91.1
28	GPT-5.6 Luna	OpenAI	91.1
28	Kimi K2.6	Moonshot AI	91.1
31	DeepSeek V4 Flash 0731	DeepSeek	90.8
31	Gemini 3 Pro	Google	90.8
33	Qwen3.8-27B	Alibaba	90.5
34	Muse Spark 1.2	Meta	90.4
35	GPT-5.2	OpenAI	90.3
36	Grok 4.3	xAI	90.1
37	Qwen3.7 Plus	Alibaba	90.0
38	GPT-5.2-Codex	OpenAI	89.9
39	Muse Spark 1.1	Meta	89.8
40	Hy3	Tencent	89.7
40	Hy3 Preview	Tencent	89.7
42	Claude Opus 4.6 (Adaptive)	Anthropic	89.6
42	Kimi K2.7 Code	Moonshot AI	89.6
44	GLM-5.2	Z.AI	89.5
44	TMInkling-Small	Thinking Machines Lab	89.5
46	Solar Pro 4	Upstage	89.1
47	Qwen 3.6 Max (preview)	Alibaba	88.8
48	Claude Opus 4.7	Anthropic	88.5
49	Muse Spark	Meta	88.4
50	Qwen3.6 Plus	Alibaba	88.2
51	Kimi K2.5	Moonshot AI	87.9
51	Kimi K2.5 (Reasoning)	Moonshot AI	87.9
53	Grok 4	xAI	87.7
54	GPT-5.4 mini	OpenAI	87.5
55	MiniMax M2.7	MiniMax	87.4
56	GPT-5.1	OpenAI	87.3
57	TMInkling	Thinking Machines Lab	87.2
58	MiMo-V2-Pro	Xiaomi	87.0
59	GLM-5.1	Z.AI	86.8
60	Nemotron 3 Ultra	NVIDIA	86.7
61	Claude Opus 4.5 Thinking	Anthropic	86.6
61	MiMo-V2.5-Pro	Xiaomi	86.6
63	APApodex 1.1	Apodex	86.4
63	APApodex 1.1 Mini	Apodex	86.4
65	Ling 3.0 Flash VL	InclusionAI	86.2
66	Qwen3.5 397B	Alibaba	86.1
66	Qwen3.5 397B (Reasoning)	Alibaba	86.1
68	GPT-5.1-Codex	OpenAI	86.0
68	GPT-5.1-Codex-Max	OpenAI	86.0
70	GLM-4.7	Z.AI	85.9
71	Qwen3.5-27B	Alibaba	85.8
72	STA.X K2	SK Telecom	85.7
72	Gemma 4 31B	Google	85.7
72	Qwen3.5-122B-A10B	Alibaba	85.7
75	Ling 3.0 Flash	InclusionAI	85.5
75	Ling 3.0 Flash FP8	InclusionAI	85.5
77	GPT-5 (high)	OpenAI	85.4
78	Grok 4.1 Fast (Reasoning)	xAI	85.3
79	GLM-5-Turbo	Z.AI	84.7
79	Grok 4 Fast (Reasoning)	xAI	84.7
81	o3-pro	OpenAI	84.5
81	Qwen3.5-35B-A3B	Alibaba	84.5
83	Gemini 2.5 Pro	Google	84.4
84	GPT-5 (medium)	OpenAI	84.2
84	Qwen3.6-27B	Alibaba	84.2
86	Qwen3.6-35B-A3B	Alibaba	84.1
87	Claude Opus 4.6	Anthropic	84.0
88	Gemini 3.5 Flash-Lite	Google	83.8
89	Muse Glimmer 30B	Meta	83.5
90	K-EXAONE 2.0	LG AI Research	82.9
91	MiMo-V2-Omni	Xiaomi	82.8
92	o3	OpenAI	82.7
93	GLM-5	Z.AI	82.0
94	GPT-5.4 nano	OpenAI	81.7
95	DeepSeek-R1	DeepSeek	81.3
96	Gemini 3 Flash	Google	81.2
97	Claude Opus 4.5	Anthropic	81.0
98	Claude 4.1 Opus Thinking	Anthropic	80.9
98	GLM-5V-Turbo	Z.AI	80.9
98	Step 3.7 Flash	StepFun	80.9
101	Nemotron 3 Super 100B	NVIDIA	80.0
102	Claude Sonnet 4.6	Anthropic	79.9
103	Gemma 4 26B A4B	Google	79.2
104	K-Exaone	LG AI Research	78.3
105	GPT-OSS 120B	OpenAI	78.2
106	DeepSeek V3.1 (Reasoning)	DeepSeek	77.9
107	Mistral Small 4	Mistral	76.9
107	Mistral Small 4 (Reasoning)	Mistral	76.9
109	Kimi K2	Moonshot AI	76.6
110	o1-preview	OpenAI	76.5
111	Qwen3 Max	Alibaba	76.4
112	Command A+	Cohere	76.1
113	Nemotron 3 Nano 30B	NVIDIA	75.7
113	North Mini Code	Cohere	75.7
115	Gemma 4 12B	Google	75.3
116	Trinity-Large-Preview	Arcee AI	75.2
116	Trinity-Large-Thinking	Arcee AI	75.2
118	DeepSeek V3.2	DeepSeek	75.1
119	Mistral Medium 3.5 128B	Mistral	74.8
119	o3-mini	OpenAI	74.8
121	o1	OpenAI	74.7
122	Nemotron 3.5 Lightning 30B A3B NVFP4	NVIDIA	74.3
123	SASarvam 105B	Sarvam	73.8
124	DeepSeek V3.1	DeepSeek	73.5
125	Ling 3.0 Tiny	InclusionAI	73.4
126	GLM-4.5-Air	Z.AI	73.3
127	MCQuasar 438B	Multiverse Computing	73.2
128	Nemotron Ultra 253B	NVIDIA	72.8
129	Grok Code Fast 1	xAI	72.7
130	Qwen3-Omni-30B-A3B-Thinking	Alibaba	72.6
131	Solar Pro 3	Upstage	72.4
132	OPMiniCPM5-2B	OpenBMB	70.2
133	GPT-OSS 20B	OpenAI	68.8
134	Claude 4 Sonnet	Anthropic	68.3
134	Gemini 2.5 Flash	Google	68.3
136	Mistral Large 3	Mistral	68.0
137	Llama 4 Maverick	Meta	67.1
138	GPT-4.1	OpenAI	66.6
139	GPT-4.1 mini	OpenAI	66.4
140	MiMo-V2-Flash	Xiaomi	65.6
141	Granite 4.2 30B	IBM	64.4
142	Grok 4.1 Fast	xAI	63.7
143	SASarvam 30B	Sarvam	63.3
144	GLM-4.6	Z.AI	63.2
145	CECeleris-1	Celeris	63.1
145	Granite 4.2 8B	IBM	63.1
147	Exaone 4.0 32B	LG AI Research	62.8
148	Qwen3-Omni-30B-A3B-Instruct	Alibaba	62.0
149	DeepSeek R1 Distill Qwen 32B	DeepSeek	61.5
150	Ling 2.6 Flash	InclusionAI	59.3
151	Gemini 1.5 Pro	Google	58.9
152	Llama 4 Scout	Meta	58.7
153	Mistral Medium 3	Mistral	57.8
154	Gemma 4 E4B	Google	57.6
155	Phi-4	Microsoft	57.5
156	Solar Pro 2	Upstage	56.1
157	Granite 4.2 3B	IBM	55.9
158	LFM2.5-2.6B	LiquidAI	55.8
159	DeepSeek V3	DeepSeek	55.7
160	GPT-4o	OpenAI	54.3
161	Llama 3.1 405B	Meta	51.5
162	LFM2.5-8B-A1B	LiquidAI	51.3
163	GPT-4.1 nano	OpenAI	51.2
164	Nova Pro	Amazon	49.9
165	FAUltravox v0.6 Llama 3.3 70B	Fixie AI	49.8
166	Claude 3 Opus	Anthropic	48.9
167	Mistral Large 2	Mistral	48.6
168	Nemotron 3 Nano Omni 30B A3B	NVIDIA	46.9
169	Gemma 4 E2B	Google	43.3
170	Gemma 3 27B	Google	42.8
171	GPT-4o mini	OpenAI	42.6
172	Exaone 4.0 1.2B	LG AI Research	42.4
173	Qwen2.5 Coder 32B Instruct	Alibaba	41.7
174	Claude 3 Haiku	Anthropic	37.4
175	Phi-4 Multimodal Instruct	Microsoft	31.5
176	LFM2.5-VL-1.6B-Extract	LiquidAI	28.9
177	Granite-4.0-1B	IBM	28.1
178	Gemini 1.0 Pro	Google	27.7
179	Granite-4.0-H-1B	IBM	26.3
180	Granite-4.0-350M	IBM	26.1
181	Granite-4.0-H-350M	IBM	25.7

Evidence key: ObservedLast good

Rows are ordered by the value Artificial Analysis GPQA Diamond Benchmark Leaderboard 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.