Humanity's Last Exam without tools

Tool-free variant of Humanity's Last Exam that isolates a model's raw frontier reasoning.

In index
Knowledge

Models38

Top result64.4%Claude Opus 5.5

Top-3 spread5.4 pts

YearN/A

Result and price

Results

38 results

#ModelResult

01

Claude Opus 5.564.4%

02

Claude Fable 5.160.9%

03

Claude Mythos 559.0%

04

Claude Opus 556.3%

05

Muse Spark 1.152.2%

06SASakana Fugu-Ultra50.0%

07

Claude Opus 4.849.8%

08SASakana Fugu47.2%

09

Claude Opus 4.7 (Adaptive)46.9%

10

Gemini 3.1 Pro45.4%

11OAOrnith-1.5-397B44.6%

12

Qwen3.8 Max43.6%

13MAKimi K343.5%

14TEHy4 preview43.4%

15

Claude Sonnet 543.2%

16

GPT-5.5 Pro43.1%

17

Muse Spark42.8%

18

GPT-5.4 Pro42.7%

19

GPT-5.541.4%

20

GLM-5.240.5%

21

Claude Opus 4.640.0%

22

GPT-5.439.8%

23

Qwen3.8-Flash-Next35.9%

24

MiMo-V2.5-Pro34.0%

25TMInkling-Small31.6%

25

Grok 4.2031.6%

27

Qwen3.8-27B30.8%

28TMInkling30.0%

29UPSolar Open 228.8%

30

GPT-5.4 mini28.2%

31

Nemotron 3 Ultra26.7%

32OAOrnith-1.5-35B-A3B25.6%

33

GPT-5.4 nano24.3%

34OAOrnith-1.5-9B20.2%

35

Gemma 4 31B19.5%

36

Nemotron 3.5 Lightning 30B A3B NVFP410.5%

37

Gemma 4 26B A4B8.7%

38

Gemma 4 12B5.2%