Skip to main content
BenchLM
Advertise

Benchmark profile

Artificial Analysis GPQA Diamond (AA-GPQA Diamond)

A display-only Artificial Analysis GPQA Diamond score.

Data verified

Benchmark score on AA-GPQA Diamond — July 29, 2026

BenchLM mirrors the published score view for AA-GPQA Diamond. GPT-5.6 Sol leads the public snapshot at 94.1% , followed by Gemini 3.1 Pro (94.1%) and Kimi K3 (93.5%). BenchLM does not use these results to rank models overall.

1Closed

GPT-5.6 Sol

OpenAI

gpt-5-6-sol

94.1%
Overall 81.39Context 1M
2Closed

Gemini 3.1 Pro

Google

gemini-3-1-pro

94.1%
Overall 54.66Context 1M
3Closed

Kimi K3

Moonshot AI

kimi-3

93.5%
Overall 79.89Context 1.05M
163 modelsKnowledgeCurrentDisplay onlyUpdated July 29, 2026

Benchmark score table (163 models)

Score
1
GPT-5.6 SolOpenAI · Closed
94.1%
2
Gemini 3.1 ProGoogle · Closed
94.1%
3
Kimi K3Moonshot AI · Closed
93.5%
4
GPT-5.5OpenAI · Closed
93.5%
5
Claude Opus 5Anthropic · Closed
93.2%
6
Grok 4.5xAI · Closed
93.1%
7
MiniMax M3MiniMax · Open weight
92.9%
8
Gemini 3.6 FlashGoogle · Closed
92.8%
9
Claude Fable 5Anthropic · Closed
92.6%
10
GPT-5.6 TerraOpenAI · Closed
92.5%
11
Qwen3.7 MaxAlibaba · Closed
92.3%
12
Gemini 3.5 FlashGoogle · Closed
92.2%
13
Claude Opus 4.8Anthropic · Closed
92.0%
14
GPT-5.4OpenAI · Closed
92.0%
15
GPT-5.3 CodexOpenAI · Closed
91.5%
16
GPT-5.3-Codex-SparkOpenAI · Closed
91.5%
17
Claude Opus 4.7 (Adaptive)Anthropic · Closed
91.4%
18
Claude Sonnet 5Anthropic · Closed
91.1%
19
GPT-5.6 LunaOpenAI · Closed
91.1%
20
Kimi K2.6Moonshot AI · Open weight
91.1%
21
Gemini 3 ProGoogle · Closed
90.8%
22
DeepSeek V4 Pro (High)DeepSeek · Open weight
90.5%
23
GPT-5.2OpenAI · Closed
90.3%
24
Grok 4.3xAI · Closed
90.1%
25
Qwen3.7 PlusAlibaba · Closed
90.0%
26
GPT-5.2-CodexOpenAI · Closed
89.9%
27
Muse Spark 1.1Meta · Closed
89.8%
28
Hy3 PreviewTencent · Open weight
89.7%
29
Hy3Tencent · Open weight
89.7%
30
Claude Opus 4.6 (Adaptive)Anthropic · Closed
89.6%
31
Kimi K2.7 CodeMoonshot AI · Open weight
89.6%
32
GLM-5.2Z.AI · Open weight
89.5%
33
DeepSeek V4 Flash (Max)DeepSeek · Open weight
89.4%
34
Qwen3.5 397BAlibaba · Open weight
89.3%
35
Qwen3.5 397B (Reasoning)Alibaba · Open weight
89.3%
36
DeepSeek V4 Pro (Max)DeepSeek · Open weight
88.8%
37
Qwen 3.6 Max (preview)Alibaba · Closed
88.8%
38
Claude Opus 4.7Anthropic · Closed
88.5%
39
Muse SparkMeta · Closed
88.4%
40
Qwen3.6 PlusAlibaba · Closed
88.2%
41
Kimi K2.5Moonshot AI · Open weight
87.9%
42
Kimi K2.5 (Reasoning)Moonshot AI · Closed
87.9%
43
Grok 4xAI · Closed
87.7%
44
GPT-5.4 miniOpenAI · Closed
87.5%
45
MiniMax M2.7MiniMax · Open weight
87.4%
46
GPT-5.1OpenAI · Closed
87.3%
47
InklingThinking Machines Lab · Open weight
87.2%
48
MiMo-V2-ProXiaomi · Closed
87.0%
49
GLM-5.1Z.AI · Open weight
86.8%
50
Nemotron 3 UltraNVIDIA · Open weight
86.7%
51
DeepSeek V4 Flash (High)DeepSeek · Open weight
86.7%
52
MiMo-V2.5-ProXiaomi · Closed
86.6%
53
Claude Opus 4.5 ThinkingAnthropic · Closed
86.6%
54
GPT-5.1-Codex-MaxOpenAI · Closed
86.0%
55
GPT-5.1-CodexOpenAI · Closed
86.0%
56
GLM-4.7Z.AI · Open weight
85.9%
57
Qwen3.5-27BAlibaba · Open weight
85.8%
58
Qwen3.5-122B-A10BAlibaba · Open weight
85.7%
59
Gemma 4 31BGoogle · Open weight
85.7%
60
GPT-5 (high)OpenAI · Closed
85.4%
61
Grok 4.1 Fast (Reasoning)xAI · Closed
85.3%
62
MiniMax M2.5MiniMax · Closed
84.8%
63
GLM-5-TurboZ.AI · Closed
84.7%
64
Grok 4 Fast (Reasoning)xAI · Closed
84.7%
65
Qwen3.5-35B-A3BAlibaba · Open weight
84.5%
66
o3-proOpenAI · Closed
84.5%
67
Gemini 2.5 ProGoogle · Closed
84.4%
68
Qwen3.6-27BAlibaba · Open weight
84.2%
69
GPT-5 (medium)OpenAI · Closed
84.2%
70
Qwen3.6-35B-A3BAlibaba · Open weight
84.1%
71
Claude Opus 4.6Anthropic · Closed
84.0%
72
Gemini 3.5 Flash-LiteGoogle · Closed
83.8%
73
MiMo-V2-OmniXiaomi · Closed
82.8%
74
GPT-5 miniOpenAI · Closed
82.8%
75
o3OpenAI · Closed
82.7%
76
Step 3.5 FlashStepFun · Open weight
82.6%
77
Gemini 3.1 Flash-LiteGoogle · Closed
82.2%
78
GLM-5Z.AI · Open weight
82.0%
79
GPT-5.4 nanoOpenAI · Closed
81.7%
80
DeepSeek-R1DeepSeek · Open weight
81.3%
81
Gemini 3 FlashGoogle · Closed
81.2%
82
Claude Opus 4.5Anthropic · Closed
81.0%
83
Step 3.7 FlashStepFun · Open weight
80.9%
84
GLM-5V-TurboZ.AI · Closed
80.9%
85
Claude 4.1 Opus ThinkingAnthropic · Closed
80.9%
86
Nemotron 3 Super 120B A12BNVIDIA · Open weight
80.0%
87
Claude Sonnet 4.6Anthropic · Closed
79.9%
88
Gemma 4 26B A4BGoogle · Open weight
79.2%
89
K-ExaoneLG AI Research · Closed
78.3%
90
GPT-OSS 120BOpenAI · Open weight
78.2%
91
DeepSeek V3.1 (Reasoning)DeepSeek · Open weight
77.9%
92
Mercury 2Inception · Closed
77.0%
93
Mistral Small 4Mistral · Open weight
76.9%
94
Mistral Small 4 (Reasoning)Mistral · Open weight
76.9%
95
Kimi K2Moonshot AI · Closed
76.6%
96
Qwen3 MaxAlibaba · Closed
76.4%
97
Command A+Cohere · Open weight
76.1%
98
Nemotron 3 Nano 30BNVIDIA · Open weight
75.7%
99
Gemma 4 12BGoogle · Open weight
75.3%
100
Trinity-Large-PreviewArcee AI · Open weight
75.2%
101
Trinity-Large-ThinkingArcee AI · Open weight
75.2%
102
DeepSeek V3.2DeepSeek · Open weight
75.1%
103
Mistral Medium 3.5 128BMistral · Open weight
74.8%
104
o3-miniOpenAI · Closed
74.8%
105
o1OpenAI · Closed
74.7%
106
Sarvam 105BSarvam · Open weight
73.8%
107
DeepSeek V3.1DeepSeek · Open weight
73.5%
108
GLM-4.5-AirZ.AI · Closed
73.3%
109
Nemotron Ultra 253BNVIDIA · Open weight
72.8%
110
Grok Code Fast 1xAI · Closed
72.7%
111
MiniMax M1 80kMiniMax · Closed
69.7%
112
GPT-OSS 20BOpenAI · Open weight
68.8%
113
Claude 4 SonnetAnthropic · Closed
68.3%
114
Gemini 2.5 FlashGoogle · Closed
68.3%
115
Mistral Large 3Mistral · Closed
68.0%
116
GPT-5 nanoOpenAI · Closed
67.6%
117
Llama 4 MaverickMeta · Open weight
67.1%
118
GPT-4.1OpenAI · Closed
66.6%
119
GPT-4.1 miniOpenAI · Closed
66.4%
120
MiMo-V2-FlashXiaomi · Open weight
65.6%
121
Grok 4.1 FastxAI · Closed
63.7%
122
Sarvam 30BSarvam · Open weight
63.3%
123
GLM-4.6Z.AI · Open weight
63.2%
124
Exaone 4.0 32BLG AI Research · Open weight
62.8%
125
DeepSeek R1 Distill Qwen 32BDeepSeek · Open weight
61.5%
126
Ling 2.6 FlashInclusionAI · Open weight
59.3%
127
Gemini 1.5 ProGoogle · Closed
58.9%
128
Llama 4 ScoutMeta · Open weight
58.7%
129
GLM-4.7-FlashZ.AI · Open weight
58.1%
130
Mistral Medium 3Mistral · Closed
57.8%
131
Phi-4Microsoft · Open weight
57.5%
132
Ministral 3 14B (Reasoning)Mistral · Open weight
57.2%
133
Ministral 3 14BMistral · Open weight
57.2%
134
Solar Pro 2Upstage · Closed
56.1%
135
DeepSeek V3DeepSeek · Open weight
55.7%
136
GPT-4oOpenAI · Closed
54.3%
137
Gemma 4 E4BGoogle · Open weight
52.2%
138
Llama 3.1 405BMeta · Open weight
51.5%
139
GPT-4.1 nanoOpenAI · Closed
51.2%
140
Nova ProAmazon · Closed
49.9%
141
Claude 3 OpusAnthropic · Closed
48.9%
142
Mistral Large 2Mistral · Closed
48.6%
143
LFM2-24B-A2BLiquidAI · Closed
47.4%
144
Ministral 3 8B (Reasoning)Mistral · Open weight
47.1%
145
Ministral 3 8BMistral · Open weight
47.1%
146
Nemotron 3 Nano Omni 30B A3BNVIDIA · Open weight
46.9%
147
LFM2.5-8B-A1BLiquidAI · Open weight
46.6%
148
Gemma 3 27BGoogle · Open weight
42.8%
149
GPT-4o miniOpenAI · Closed
42.6%
150
Exaone 4.0 1.2BLG AI Research · Open weight
42.4%
151
Qwen2.5 Coder 32B InstructAlibaba · Open weight
41.7%
152
Gemma 4 E2BGoogle · Open weight
40.5%
153
Claude 3 HaikuAnthropic · Closed
37.4%
154
LFM2.5-1.2B-ThinkingLiquidAI · Closed
33.9%
155
Ministral 3 3B (Reasoning)Mistral · Open weight
30.1%
156
Ministral 3 3BMistral · Open weight
30.1%
157
LFM2.5-1.2B-InstructLiquidAI · Closed
29.5%
158
Granite-4.0-H-350MIBM · Open weight
29.0%
159
LFM2.5-VL-1.6B-ExtractLiquidAI · Open weight
28.9%
160
Granite-4.0-1BIBM · Open weight
28.1%
161
Gemini 1.0 ProGoogle · Closed
27.7%
162
Granite-4.0-H-1BIBM · Open weight
24.6%
163
Granite-4.0-350MIBM · Open weight
23.7%

The published AA-GPQA Diamond snapshot places GPT-5.6 Sol first at 94.1%. The third row is 0.6 points behind. The broader top-10 range is 1.6 points, so many of the published results sit in a relatively narrow band.

163 models have been evaluated on AA-GPQA Diamond. The benchmark falls in the Knowledge category. This category carries a 12% weight in BenchLM.ai's overall scoring system. AA-GPQA Diamond is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.

About AA-GPQA Diamond

Year

2026

Tasks

Graduate-level science questions

Format

Accuracy

Difficulty

Graduate-level science reasoning

BenchLM stores the Artificial Analysis GPQA Diamond result separately from the weighted GPQA lane so AA refreshes remain display-only.

Artificial Analysis GPQA Diamond Benchmark Leaderboard

BenchLM freshness & provenance

Version

AA-GPQA Diamond 2026

Refresh cadence

Quarterly

Staleness state

Current

Question availability

Public benchmark set

CurrentDisplay only

BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.

FAQ

What does AA-GPQA Diamond measure?

A display-only Artificial Analysis GPQA Diamond score.

Which model scores highest on AA-GPQA Diamond?

GPT-5.6 Sol by OpenAI currently leads with a score of 94.1% on AA-GPQA Diamond.

How many models are evaluated on AA-GPQA Diamond?

163 AI models have been evaluated on AA-GPQA Diamond on BenchLM.

Compare Top Models on AA-GPQA Diamond

GPT-5.6 Sol vs Gemini 3.1 ProGemini 3.1 Pro vs Kimi K3Kimi K3 vs GPT-5.5GPT-5.5 vs Claude Opus 5
Last updated: July 29, 2026 · BenchLM version AA-GPQA Diamond 2026

Know when it’s worth switching models

The model to choose, the cheaper alternative, and the release we would wait on.

Read a sample issue

Join 2,000+ readers.

One email each week. Unsubscribe anytime.