BenchLeader
Menu
Best LLMsChoose a modelLeaderboardsModelsBenchmarksProvidersReleases9

Humanity's Last Exam (AA)

Artificial Analysis' own HLE run. Shown for reference.

As of 25 Oct 2026, Claude Opus 5.5 leads Humanity's Last Exam (AA) on BenchLeader with 61.4%, ahead of Claude Fable 5.1 at 59.1%, across 615 model configurations with a published result.

Published by
Artificial Analysis
Category
Reasoning
Index weight
Reference only
Models
615
Data as of
25 Oct 2026
Newest result
23 Oct 2026

Source: Artificial Analysis (artificialanalysis.ai). Data taken from the public leaderboard.

What the test looks like

Humanity's Last Exam is about 2,500 questions written by subject experts across more than a hundred fields, from classics to chemistry, chosen because frontier models got them wrong at the time of writing. Most are exact-answer or multiple-choice; some include images.

How it is scored

Accuracy on the exact-answer and multiple-choice questions.

What to keep in mind

Artificial Analysis' own run of Humanity's Last Exam, text-only questions, in its fixed harness. Shown for reference; the official Scale AI and CAIS board is counted in the index. Artificial Analysis omits the image questions, so scores differ slightly. See also the Humanity's Last Exam board.

Top 15
  1. 1Claude Opus 5.5 (max)61.4%
  2. 2Claude Fable 5.1 (max)59.1%
  3. 3Claude Fable 5.1 (xhigh)58.7%
  4. 4Claude Opus 5.5 (xhigh)57.5%
  5. 5Gemini 4 Argon (high)57.1%
  6. 6Claude Fable 5.1 (high)55.9%
  7. 7Claude Opus 5.5 (high)55.6%
  8. 8Claude Fable 5 (max)55.5%
  9. 9Claude Sonnet 5.5 (max)55.0%
  10. 10Claude Opus 5 (max)54.9%
  11. 11Claude Opus 5.5 (medium)54.7%
  12. 12GPT-6 Astra (max)54.7%
  13. 13GPT-6 Astra (xhigh)54.6%
  14. 14Claude Opus 5 (xhigh)54.4%
  15. 15Claude Fable 5.1 (medium)53.8%
615 of 615
#
1Claude Opus 5.5max61.4%72.62026-09-22
2Claude Fable 5.1max59.1%70.52026-09-01
3Claude Fable 5.1xhigh58.7%69.52026-09-01
4Claude Opus 5.5xhigh57.5%70.02026-09-22
5Gemini 4 Argonhigh57.1%70.72026-09-30
6Claude Fable 5.1high55.9%69.42026-09-01
7Claude Opus 5.5high55.6%69.42026-09-22
8Claude Fable 5max55.5%69.32026-06-09
9Claude Sonnet 5.5max55.0%69.42026-09-28
10Claude Opus 5max54.9%68.82026-07-24
11Claude Opus 5.5medium54.7%66.52026-09-22
12GPT-6 Astramax54.7%70.42026-09-03
13GPT-6 Astraxhigh54.6%68.92026-09-03
14Claude Opus 5xhigh54.4%67.42026-07-24
15Claude Fable 5.1medium53.8%67.32026-09-01
16Muse Wonder Fullmax53.6%70.62026-10-12
17GPT-6 Astrahigh53.1%69.62026-09-03
18GPT-6.1 Solmax52.9%69.92026-09-29
19Claude Opus 5high52.8%67.92026-07-24
20GPT-6 Astramedium52.7%67.52026-09-03
21GPT-6.1 Solxhigh52.6%67.72026-09-29
22NewKimi K3.1 Pro52.5%66.92026-10-23
23GPT-6.1 Solhigh51.4%67.72026-09-29
24Claude Opus 5medium51.3%64.62026-07-24
25Muse Wondermax50.1%67.52026-10-12
26Claude Sonnet 5.5xhigh50.0%67.92026-09-28
27GPT-6.1 Solmedium49.9%67.22026-09-29
28GPT-5.6 Solmax49.5%68.02026-07-09
29MiMo-V2.6-Pro49.4%64.02026-09-21
30GPT-6 Astralow49.2%65.92026-09-03
31Claude Fable 5.1low48.9%65.02026-09-01
32Muse Spark 1.3max48.7%67.12026-09-02
33Claude Opus 4.8max48.7%63.32026-05-28
34Claude Opus 5.5low48.3%64.72026-09-22
35NewMiniMax M3.1 Flash48.0%61.92026-10-19
36GPT-6 Solmax47.9%66.42026-09-22
37Gemini 3.7 Flashhigh47.9%63.52026-08-13
38Gemini 3.8 Flashhigh47.8%64.42026-09-02
39Muse Spark 1.3xhigh47.5%64.92026-09-02
40GPT-6.1 Sollow47.4%64.92026-09-29
41GPT-5.6 Solxhigh47.3%66.62026-07-09
42Gemini 3.1 Pro47.0%60.52026-02-19
43Kimi K3max46.9%64.82026-07-16
44Step 5 Preview46.5%60.52026-09-18
45GPT-6 Solxhigh46.3%64.32026-09-22
46Muse Spark 1.1xhigh46.2%61.62026-07-09
47GPT-5.6 Solhigh46.0%66.32026-07-09
48Claude Sonnet 5.5high45.8%65.62026-09-28
49GPT-5.5xhigh45.8%66.12026-04-23
50Muse Spark 1.2xhigh45.5%63.22026-08-05
51GPT-5.5high45.0%65.32026-04-23
52Claude Haiku 5.5max44.4%60.52026-10-07
53GPT-6 Solhigh44.1%63.72026-09-22
54Grok 4.6xhigh44.1%62.52026-08-12
55Muse Spark Openxhigh43.9%62.02026-10-15
56GPT-5.4xhigh43.7%64.72026-03-05
57JT-4.1 Flash 236B A21Bthinking43.6%–2026-07-09
58Ling 3.1 Flash43.5%–2026-10-22
59Claude Opus 5low43.4%61.32026-07-24
60DeepSeek V4.1 Flash 1017max43.3%64.02026-10-17

More reasoning benchmarks · All LLM benchmarks

Cite as: BenchLeader, “Humanity's Last Exam (AA) leaderboard”, data as of 25 Oct 2026, https://www.benchleader.com/benchmarks/aa_hle.

Humanity's Last Exam (AA): questions

What does Humanity's Last Exam (AA) measure?
Humanity's Last Exam is about 2,500 questions written by subject experts across more than a hundred fields, from classics to chemistry, chosen because frontier models got them wrong at the time of writing. Most are exact-answer or multiple-choice; some include images. Scores are reported in percent of tasks solved; higher is better.
Which AI model leads Humanity's Last Exam (AA)?
Claude Opus 5.5 leads Humanity's Last Exam (AA) with 61.4% as of 25 Oct 2026, ahead of Claude Fable 5.1 at 59.1%.
How many models have Humanity's Last Exam (AA) results?
615 model configurations have a Humanity's Last Exam (AA) result on BenchLeader, all taken from Artificial Analysis.
Who runs Humanity's Last Exam (AA) and how often is it updated?
Humanity's Last Exam (AA) is published by Artificial Analysis. BenchLeader re-reads the published results once a day and records the date each result was published.
Does Humanity's Last Exam (AA) count toward the BenchLeader Index?
No. Humanity's Last Exam (AA) is shown for reference but left out of the composite index.