Skip to main content
BenchLM
Advertise

Benchmark profile

Artificial Analysis Humanity's Last Exam (AA-HLE)

A display-only Artificial Analysis Humanity's Last Exam score.

Data verified

Benchmark score on AA-HLE — July 23, 2026

BenchLM mirrors the published score view for AA-HLE. Claude Fable 5 leads the public snapshot at 53.3% , followed by GPT-5.6 Sol (47.2%) and Claude Opus 4.8 (45.7%). BenchLM does not use these results to rank models overall.

1Closed

Claude Fable 5

Anthropic

claude-fable-5

53.3%
Overall 83.68Context 1M+
2Closed

GPT-5.6 Sol

OpenAI

gpt-5-6-sol

47.2%
Overall 81.96Context 1M
3Closed

Claude Opus 4.8

Anthropic

claude-opus-4-8

45.7%
Overall 78.34Context 1M
162 modelsKnowledgeCurrentDisplay onlyUpdated July 23, 2026

Benchmark score table (162 models)

Score
1
Claude Fable 5Anthropic · Closed
53.3%
2
GPT-5.6 SolOpenAI · Closed
47.2%
3
Claude Opus 4.8Anthropic · Closed
45.7%
4
Muse Spark 1.1Meta · Closed
45.1%
5
Gemini 3.1 ProGoogle · Closed
44.7%
6
Kimi K3Moonshot AI · Closed
44.3%
7
GPT-5.5OpenAI · Closed
44.3%
8
GPT-5.6 TerraOpenAI · Closed
41.8%
9
GPT-5.4OpenAI · Closed
41.6%
10
Gemini 3.5 FlashGoogle · Closed
41.0%
11
Grok 4.5xAI · Closed
40.3%
12
GLM-5.2Z.AI · Open weight
40.1%
13
GPT-5.3 CodexOpenAI · Closed
39.9%
14
Muse SparkMeta · Closed
39.9%
15
GPT-5.3-Codex-SparkOpenAI · Closed
39.9%
16
Claude Sonnet 5Anthropic · Closed
39.6%
17
Claude Opus 4.7 (Adaptive)Anthropic · Closed
39.6%
18
Gemini 3.6 FlashGoogle · Closed
38.3%
19
Qwen3.7 MaxAlibaba · Closed
38.1%
20
GPT-5.6 LunaOpenAI · Closed
37.2%
21
Gemini 3 ProGoogle · Closed
37.2%
22
MiniMax M3MiniMax · Open weight
37.1%
23
Claude Opus 4.6 (Adaptive)Anthropic · Closed
36.7%
24
DeepSeek V4 Pro (Max)DeepSeek · Open weight
35.9%
25
Kimi K2.6Moonshot AI · Open weight
35.9%
26
GPT-5.2OpenAI · Closed
35.4%
27
Grok 4.3xAI · Closed
35.0%
28
MiMo-V2.5-ProXiaomi · Closed
33.8%
29
DeepSeek V4 Pro (High)DeepSeek · Open weight
33.5%
30
GPT-5.2-CodexOpenAI · Closed
33.5%
31
Qwen3.7 PlusAlibaba · Closed
33.4%
32
Kimi K2.7 CodeMoonshot AI · Open weight
32.8%
33
DeepSeek V4 Flash (Max)DeepSeek · Open weight
32.1%
34
Hy3 PreviewTencent · Open weight
31.6%
35
Hy3Tencent · Open weight
31.6%
36
Claude Opus 4.7Anthropic · Closed
31.2%
37
InklingThinking Machines Lab · Open weight
29.7%
38
Kimi K2.5Moonshot AI · Open weight
29.4%
39
Kimi K2.5 (Reasoning)Moonshot AI · Closed
29.4%
40
Qwen 3.6 Max (preview)Alibaba · Closed
28.9%
41
Claude Opus 4.5 ThinkingAnthropic · Closed
28.4%
42
MiMo-V2-ProXiaomi · Closed
28.3%
43
MiniMax M2.7MiniMax · Open weight
28.1%
44
GLM-5.1Z.AI · Open weight
28.0%
45
DeepSeek V4 Flash (High)DeepSeek · Open weight
27.8%
46
Qwen3.5 397BAlibaba · Open weight
27.3%
47
Qwen3.5 397B (Reasoning)Alibaba · Open weight
27.3%
48
GLM-5Z.AI · Open weight
27.2%
49
GPT-5.4 miniOpenAI · Closed
26.6%
50
Nemotron 3 UltraNVIDIA · Open weight
26.6%
51
GPT-5.4 nanoOpenAI · Closed
26.5%
52
GPT-5.1OpenAI · Closed
26.5%
53
GPT-5 (high)OpenAI · Closed
26.5%
54
Qwen3.6 PlusAlibaba · Closed
25.7%
55
GLM-5-TurboZ.AI · Closed
25.4%
56
GLM-4.7Z.AI · Open weight
25.1%
57
Grok 4xAI · Closed
23.9%
58
GPT-5 (medium)OpenAI · Closed
23.5%
59
Qwen3.5-122B-A10BAlibaba · Open weight
23.4%
60
GPT-5.1-Codex-MaxOpenAI · Closed
23.4%
61
GPT-5.1-CodexOpenAI · Closed
23.4%
62
Gemma 4 31BGoogle · Open weight
22.7%
63
Step 3.5 FlashStepFun · Open weight
22.6%
64
Qwen3.5-27BAlibaba · Open weight
22.2%
65
Qwen3.6-27BAlibaba · Open weight
21.6%
66
Gemini 2.5 ProGoogle · Closed
21.1%
67
Qwen3.6-35B-A3BAlibaba · Open weight
20.2%
68
o3OpenAI · Closed
20.0%
69
Step 3.7 FlashStepFun · Open weight
19.9%
70
MiMo-V2-OmniXiaomi · Closed
19.9%
71
Qwen3.5-35B-A3BAlibaba · Open weight
19.7%
72
GPT-5 miniOpenAI · Closed
19.7%
73
Nemotron 3 Super 120B A12BNVIDIA · Open weight
19.2%
74
MiniMax M2.5MiniMax · Closed
19.1%
75
Claude Opus 4.6Anthropic · Closed
18.6%
76
GPT-OSS 120BOpenAI · Open weight
18.5%
77
Gemma 4 26B A4BGoogle · Open weight
18.3%
78
Grok 4.1 Fast (Reasoning)xAI · Closed
17.6%
79
Gemini 3.5 Flash-LiteGoogle · Closed
17.5%
80
Grok 4 Fast (Reasoning)xAI · Closed
17.0%
81
Gemini 3.1 Flash-LiteGoogle · Closed
16.2%
82
GLM-5V-TurboZ.AI · Closed
15.8%
83
Mercury 2Inception · Closed
15.5%
84
DeepSeek-R1DeepSeek · Open weight
14.9%
85
Gemma 4 12BGoogle · Open weight
14.8%
86
Trinity-Large-PreviewArcee AI · Open weight
14.7%
87
Trinity-Large-ThinkingArcee AI · Open weight
14.7%
88
Gemini 3 FlashGoogle · Closed
14.1%
89
Claude Sonnet 4.6Anthropic · Closed
13.2%
90
K-ExaoneLG AI Research · Closed
13.1%
91
DeepSeek V3.1 (Reasoning)DeepSeek · Open weight
13.0%
92
Claude Opus 4.5Anthropic · Closed
12.9%
93
Mistral Medium 3.5 128BMistral · Open weight
12.8%
94
Claude 4.1 Opus ThinkingAnthropic · Closed
11.9%
95
Command A+Cohere · Open weight
11.4%
96
Qwen3 MaxAlibaba · Closed
11.1%
97
DeepSeek V3.2DeepSeek · Open weight
10.5%
98
Nemotron 3 Nano 30BNVIDIA · Open weight
10.2%
99
Sarvam 105BSarvam · Open weight
10.1%
100
GPT-OSS 20BOpenAI · Open weight
9.8%
101
Mistral Small 4Mistral · Open weight
9.5%
102
Mistral Small 4 (Reasoning)Mistral · Open weight
9.5%
103
o3-miniOpenAI · Closed
8.7%
104
GPT-5 nanoOpenAI · Closed
8.2%
105
MiniMax M1 80kMiniMax · Closed
8.2%
106
Nemotron Ultra 253BNVIDIA · Open weight
8.1%
107
MiMo-V2-FlashXiaomi · Open weight
8.0%
108
o1OpenAI · Closed
7.7%
109
Grok Code Fast 1xAI · Closed
7.5%
110
GLM-4.7-FlashZ.AI · Open weight
7.1%
111
Kimi K2Moonshot AI · Closed
7.0%
112
Sarvam 30BSarvam · Open weight
7.0%
113
LFM2.5-8B-A1BLiquidAI · Open weight
6.9%
114
GLM-4.5-AirZ.AI · Closed
6.8%
115
LFM2.5-1.2B-InstructLiquidAI · Closed
6.8%
116
Granite-4.0-H-350MIBM · Open weight
6.4%
117
DeepSeek V3.1DeepSeek · Open weight
6.3%
118
Ling 2.6 FlashInclusionAI · Open weight
6.2%
119
LFM2.5-1.2B-ThinkingLiquidAI · Closed
6.1%
120
Exaone 4.0 1.2BLG AI Research · Open weight
5.8%
121
Granite-4.0-350MIBM · Open weight
5.7%
122
DeepSeek R1 Distill Qwen 32BDeepSeek · Open weight
5.5%
123
Nemotron 3 Nano Omni 30B A3BNVIDIA · Open weight
5.3%
124
Ministral 3 3B (Reasoning)Mistral · Open weight
5.3%
125
Ministral 3 3BMistral · Open weight
5.3%
126
GLM-4.6Z.AI · Open weight
5.2%
127
Gemini 2.5 FlashGoogle · Closed
5.1%
128
LFM2.5-VL-1.6B-ExtractLiquidAI · Open weight
5.1%
129
Granite-4.0-1BIBM · Open weight
5.1%
130
Grok 4.1 FastxAI · Closed
5.0%
131
Granite-4.0-H-1BIBM · Open weight
5.0%
132
Exaone 4.0 32BLG AI Research · Open weight
4.9%
133
Gemini 1.5 ProGoogle · Closed
4.9%
134
Llama 4 MaverickMeta · Open weight
4.8%
135
Gemma 4 E2BGoogle · Open weight
4.8%
136
Gemma 3 27BGoogle · Open weight
4.7%
137
GPT-4.1OpenAI · Closed
4.6%
138
GPT-4.1 miniOpenAI · Closed
4.6%
139
Gemini 1.0 ProGoogle · Closed
4.6%
140
Ministral 3 14B (Reasoning)Mistral · Open weight
4.6%
141
Ministral 3 14BMistral · Open weight
4.6%
142
LFM2-24B-A2BLiquidAI · Closed
4.4%
143
Llama 4 ScoutMeta · Open weight
4.3%
144
Mistral Medium 3Mistral · Closed
4.3%
145
Ministral 3 8B (Reasoning)Mistral · Open weight
4.3%
146
Ministral 3 8BMistral · Open weight
4.3%
147
Llama 3.1 405BMeta · Open weight
4.2%
148
Mistral Large 3Mistral · Closed
4.1%
149
Phi-4Microsoft · Open weight
4.1%
150
Claude 4 SonnetAnthropic · Closed
4.0%
151
GPT-4o miniOpenAI · Closed
4.0%
152
Mistral Large 2Mistral · Closed
4.0%
153
GPT-4.1 nanoOpenAI · Closed
3.9%
154
Claude 3 HaikuAnthropic · Closed
3.9%
155
Qwen2.5 Coder 32B InstructAlibaba · Open weight
3.8%
156
Solar Pro 2Upstage · Closed
3.8%
157
Gemma 4 E4BGoogle · Open weight
3.7%
158
DeepSeek V3DeepSeek · Open weight
3.6%
159
Nova ProAmazon · Closed
3.4%
160
GPT-4oOpenAI · Closed
3.3%
161
GPT-4 TurboOpenAI · Closed
3.3%
162
Claude 3 OpusAnthropic · Closed
3.1%

The published AA-HLE snapshot places Claude Fable 5 first at 53.3%. The third row is 7.6 points behind. The broader top-10 range is 12.3 points, so the table still separates the published systems.

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

About AA-HLE

Year

2026

Tasks

Expert-level questions

Format

Accuracy

Difficulty

Frontier expert reasoning

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

Artificial Analysis Humanity's Last Exam Benchmark Leaderboard

BenchLM freshness & provenance

Version

AA-HLE 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-HLE measure?

A display-only Artificial Analysis Humanity's Last Exam score.

Which model scores highest on AA-HLE?

Claude Fable 5 by Anthropic currently leads with a score of 53.3% on AA-HLE.

How many models are evaluated on AA-HLE?

162 AI models have been evaluated on AA-HLE on BenchLM.

Compare Top Models on AA-HLE

Claude Fable 5 vs GPT-5.6 SolGPT-5.6 Sol vs Claude Opus 4.8Claude Opus 4.8 vs Muse Spark 1.1Muse Spark 1.1 vs Gemini 3.1 Pro
Last updated: July 23, 2026 · BenchLM version AA-HLE 2026

Choose a model with this week’s evidence

Join 2,000+ readers for ranking moves, pricing changes, and the claims that still need proof.

One email each week. Unsubscribe anytime.