AI Model Release Tracker
Multidisciplinary reasoning

Humanity's Last Examno tools

Humanity's Last Exam — extremely hard expert questions across many subjects, written so you can't just look up the answer. “No tools” means the AI answers on its own. Higher is better.

no toolstext only, no toolswith tools

Rankings

Higher is better
#1
Claude Opus 5.5
64.4%
2
Claude Fable 5.1
60.9%
3
Claude Mythos 5
59%
4
Claude Fable 5
57.8%
5
Claude Sonnet 5.5
56.9%
6
Claude Mythos Preview
56.8%
7
Claude Opus 5
56.3%
8
Kimi K3.1 ProOW
51.1%
9
Claude Opus 4.8
49.8%
10
Claude Opus 4.7
46.9%
11
Claude Haiku 5.5
45.9%
12
Gemini 3.1 Pro
44.4%
13
Kimi K3
43.5%
14
Claude Sonnet 5
43.2%
15
GPT-5.5-Pro
43.1%
16
GPT-5.4-Pro
42.7%
16
DeepSeek-V4-Pro-0813OW
42.7%
18
DeepSeek-V4.1-Flash-1017OW
42.6%
19
GPT-5.5
41.4%
20
GLM-5.2
40.5%
21
Gemini 3.5 Flash
40.2%
22
Claude Opus 4.6
40%
23
GPT-5.4
39.8%
24
Gemini 3.0 Pro
37.5%
25
DeepSeek-V4.1-FlashOW
36.8%
26
Qwen3.8-Flash-NextOW
35.9%
27
GPT-5.2
34.5%
28
Gemini 3.0 Flash
33.7%
29
Claude Sonnet 4.6
33.2%
30
Claude Opus 4.5
30.8%
30
Qwen3.8-27BOW
30.8%
32
Qwen3.5OW
28.7%
33
GPT-5.4 mini
28.2%
34
GPT-5
24.8%
34
GLM-4.7
24.8%
36
GPT-5.4 nano
24.3%
37
Qwen3.6OW
21.4%
38
Gemma 4OW
19.5%
39
Gemini 2.5 Pro
18.8%
40
Claude Sonnet 4.5
17.7%
41
GPT-5 mini
16.7%
42
Gemini 3.1 Flash-Lite
16%
43
gpt-oss-120bOW
14.9%
44
Gemini 2.5 Flash
12.1%
45
gpt-oss-20bOW
10.9%
46
Claude Haiku 4.5
10.2%
47
GPT-5 nano
8.7%
48
Gemini 2.5 Flash-Lite
5.1%

Humanity's Last Exam — frequently asked questions

Humanity's Last Exam — extremely hard expert questions across many subjects, written so you can't just look up the answer. “No tools” means the AI answers on its own. Higher is better.