Skip to main content
benchlm.ai

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

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

Benchmark score on AA-HLE — July 4, 2026

BenchLM mirrors the published score view for AA-HLE. Claude Opus 4.8 leads the public snapshot at 45.7% , followed by Gemini 3.1 Pro (44.7%) and GPT-5.5 (44.3%). BenchLM does not use these results to rank models overall.

1Closed

Claude Opus 4.8

Anthropic

45.7%
Overall 85Context 1M
2Closed

Gemini 3.1 Pro

Google

44.7%
Overall 88Context 1M
3Closed

GPT-5.5

OpenAI

44.3%
Overall 78Context 1M
132 modelsKnowledgeCurrentDisplay onlyUpdated July 4, 2026

The published AA-HLE snapshot is tightly clustered at the top: Claude Opus 4.8 sits at 45.7%, while the third row is only 1.4 points behind. The broader top-10 spread is 7.6 points, so many of the published scores sit in a relatively narrow band.

132 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.

Benchmark score table (132 models)

1
Claude Opus 4.8
45.7%
2
Gemini 3.1 Pro
44.7%
3
GPT-5.5
44.3%
4
GPT-5.4
41.6%
5
Gemini 3.5 Flash
41.0%
6
GLM-5.2
40.1%
7
GPT-5.3 Codex
39.9%
8
Muse Spark
39.9%
9
Claude Opus 4.7 (Adaptive)
39.6%
10
Qwen3.7 Max
38.1%
11
Gemini 3 Pro
37.2%
12
MiniMax M3
37.1%
13
Claude Opus 4.6 (Adaptive)
36.7%
14
DeepSeek V4 Pro (Max)
35.9%
15
Kimi K2.6
35.9%
16
GPT-5.2
35.4%
17
Grok 4.3
35.0%
18
MiMo-V2.5-Pro
33.8%
19
DeepSeek V4 Pro (High)
33.5%
20
GPT-5.2-Codex
33.5%
21
Qwen3.7 Plus
33.4%
22
Kimi K2.7 Code
32.8%
23
DeepSeek V4 Flash (Max)
32.1%
24
Claude Opus 4.7
31.2%
25
Kimi K2.5 (Reasoning)
29.4%
26
Kimi K2.5
29.4%
27
Qwen 3.6 Max (preview)
28.9%
28
Claude Opus 4.5 Thinking
28.4%
29
MiMo-V2-Pro
28.3%
30
MiniMax M2.7
28.1%
31
GLM-5.1
28.0%
32
DeepSeek V4 Flash (High)
27.8%
33
Qwen3.5 397B (Reasoning)
27.3%
34
GLM-5
27.2%
35
Nemotron 3 Ultra
26.6%
36
GPT-5.4 mini
26.6%
37
GPT-5.1
26.5%
38
GPT-5 (high)
26.5%
39
GPT-5.4 nano
26.5%
40
Qwen3.6 Plus
25.7%
41
Hy3 Preview
25.5%
42
GLM-5-Turbo
25.4%
43
GLM-4.7
25.1%
44
Grok 4
23.9%
45
GPT-5 (medium)
23.5%
46
GPT-5.1-Codex-Max
23.4%
47
Qwen3.5-122B-A10B
23.4%
48
GPT-5.1-Codex
23.4%
49
Gemma 4 31B
22.7%
50
Qwen3.5-27B
22.2%
51
Qwen3.6-27B
21.6%
52
Gemini 2.5 Pro
21.1%
53
Qwen3.6-35B-A3B
20.2%
54
o3
20.0%
55
MiMo-V2-Omni
19.9%
56
Step 3.7 Flash
19.9%
57
Qwen3.5-35B-A3B
19.7%
58
Qwen3.5 397B
18.8%
59
Claude Opus 4.6
18.6%
60
GPT-OSS 120B
18.5%
61
Gemma 4 26B A4B
18.3%
62
Grok 4.1 Fast (Reasoning)
17.6%
63
Grok 4 Fast (Reasoning)
17.0%
64
Gemini 3.1 Flash-Lite
16.2%
65
GLM-5V-Turbo
15.8%
66
DeepSeek-R1
14.9%
67
Gemma 4 12B
14.8%
68
Trinity-Large-Thinking
14.7%
69
Trinity-Large-Preview
14.7%
70
Gemini 3 Flash
14.1%
71
Claude Sonnet 4.6
13.2%
72
K-Exaone
13.1%
73
DeepSeek V3.1 (Reasoning)
13.0%
74
Claude Opus 4.5
12.9%
75
Mistral Medium 3.5 128B
12.8%
76
Claude 4.1 Opus Thinking
11.9%
77
Command A+
11.4%
78
Qwen3 Max
11.1%
79
DeepSeek V3.2
10.5%
80
Sarvam 105B
10.1%
81
GPT-OSS 20B
9.8%
82
Mistral Small 4 (Reasoning)
9.5%
83
Mistral Small 4
9.5%
84
o3-mini
8.7%
85
Nemotron Ultra 253B
8.1%
86
MiMo-V2-Flash
8.0%
87
o1
7.7%
88
Grok Code Fast 1
7.5%
89
Kimi K2
7.0%
90
Sarvam 30B
7.0%
91
LFM2.5-8B-A1B
6.9%
92
GLM-4.5-Air
6.8%
93
Granite-4.0-H-350M
6.4%
94
DeepSeek V3.1
6.3%
95
Ling 2.6 Flash
6.2%
96
Exaone 4.0 1.2B
5.8%
97
Granite-4.0-350M
5.7%
98
DeepSeek R1 Distill Qwen 32B
5.5%
99
Nemotron 3 Nano Omni 30B A3B
5.3%
100
GLM-4.6
5.2%
101
Gemini 2.5 Flash
5.1%
102
Granite-4.0-1B
5.1%
103
LFM2.5-VL-1.6B-Extract
5.1%
104
Grok 4.1 Fast
5.0%
105
Granite-4.0-H-1B
5.0%
106
Gemini 1.5 Pro
4.9%
107
Exaone 4.0 32B
4.9%
108
Llama 4 Maverick
4.8%
109
Gemma 4 E2B
4.8%
110
Gemma 3 27B
4.7%
111
GPT-4.1
4.6%
112
GPT-4.1 mini
4.6%
113
Nemotron 3 Nano 30B
4.6%
114
Gemini 1.0 Pro
4.6%
115
Llama 4 Scout
4.3%
116
Mistral Medium 3
4.3%
117
Llama 3.1 405B
4.2%
118
Mistral Large 3
4.1%
119
Phi-4
4.1%
120
Claude 4 Sonnet
4.0%
121
GPT-4o mini
4.0%
122
Mistral Large 2
4.0%
123
GPT-4.1 nano
3.9%
124
Claude 3 Haiku
3.9%
125
Qwen2.5 Coder 32B Instruct
3.8%
126
Solar Pro 2
3.8%
127
Gemma 4 E4B
3.7%
128
DeepSeek V3
3.6%
129
Nova Pro
3.4%
130
GPT-4o
3.3%
131
GPT-4 Turbo
3.3%
132
Claude 3 Opus
3.1%

FAQ

What does AA-HLE measure?

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

Which model scores highest on AA-HLE?

Claude Opus 4.8 by Anthropic currently leads with a score of 45.7% on AA-HLE.

How many models are evaluated on AA-HLE?

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

Compare Top Models on AA-HLE

Claude Opus 4.8 vs Gemini 3.1 ProGemini 3.1 Pro vs GPT-5.5GPT-5.5 vs GPT-5.4GPT-5.4 vs Gemini 3.5 Flash
Last updated: July 4, 2026 · BenchLM version AA-HLE 2026

The AI models change fast. We track them for you.

For engineers, researchers, and the plain curious — a weekly brief on new models, ranking shifts, and pricing changes.

Free. No spam. Unsubscribe anytime.