BenchLeader
Menu
Best LLMsChoose a modelLeaderboardsModelsBenchmarksProvidersReleases9

GPQA Diamond (AA)

Artificial Analysis' own GPQA Diamond run. Shown for reference; Epoch's run is in the index.

As of 26 Oct 2026, GPT-6 Astra leads GPQA Diamond (AA) on BenchLeader with 96.3%, ahead of GPT-6 Astra at 96.1%, across 559 model configurations with a published result.

Published by
Artificial Analysis
Category
Reasoning
Index weight
Reference only
Models
559
Data as of
26 Oct 2026
Newest result
15 Oct 2026

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

What the test looks like

GPQA Diamond is a set of 198 multiple-choice questions in biology, physics and chemistry written by PhD-level domain experts. The questions are designed to be "Google-proof": a skilled non-expert with unlimited web access still answers only about a third correctly, while experts in the field score around two-thirds. The Diamond subset keeps only the questions that experts answered correctly and non-experts got wrong.

How it is scored

Accuracy on the 198 questions.

What to keep in mind

Artificial Analysis' own run of GPQA Diamond in its fixed harness. Shown for reference; Epoch AI's run is the one counted in the index. Differences between the two runs show how much prompting and sampling matter. See also the GPQA Diamond board.

Top 15
  1. 1GPT-6 Astra (xhigh)96.3%
  2. 2GPT-6 Astra (max)96.1%
  3. 3Gemini 3.8 Flash (high)95.3%
  4. 4GPT-6 Astra (high)95.0%
  5. 5Grok 4.6 (high)95.0%
  6. 6Gemini 3.7 Flash (high)94.5%
  7. 7GPT-5.6 Sol (max)94.1%
  8. 8Muse Spark 1.3 (xhigh)94.1%
  9. 9Gemini 3.1 Pro94.1%
  10. 10GPT-6 Astra (medium)93.9%
  11. 11Claude Fable 5.1 (max)93.7%
  12. 12Claude Opus 5 (high)93.7%
  13. 13Claude Opus 5 (xhigh)93.7%
  14. 14Muse Spark 1.3 (max)93.5%
  15. 15GPT-5.5 (xhigh)93.5%
559 of 559
#
1GPT-6 Astraxhigh96.3%68.92026-09-03
2GPT-6 Astramax96.1%70.42026-09-03
3Gemini 3.8 Flashhigh95.3%64.42026-09-02
4GPT-6 Astrahigh95.0%69.62026-09-03
5Grok 4.6high95.0%62.92026-08-12
6Gemini 3.7 Flashhigh94.5%63.52026-08-13
7GPT-5.6 Solmax94.1%68.02026-07-09
8Muse Spark 1.3xhigh94.1%64.92026-09-02
9Gemini 3.1 Pro94.1%61.82026-02-19
10GPT-6 Astramedium93.9%67.52026-09-03
11Claude Fable 5.1max93.7%70.52026-09-01
12Claude Opus 5high93.7%67.92026-07-24
13Claude Opus 5xhigh93.7%67.42026-07-24
14Muse Spark 1.3max93.5%67.12026-09-02
15GPT-5.5xhigh93.5%66.12026-04-23
16Kimi K3max93.5%64.82026-07-16
17Gemini 3.8 Flashmedium93.5%64.42026-09-02
18Grok 4.6medium93.5%64.22026-08-12
19Grok 4.6xhigh93.5%62.52026-08-12
20Qwen3.8 2.4T A95B93.5%61.02026-08-12
21Claude Fable 5.1xhigh93.4%69.52026-09-01
22Claude Opus 5max93.2%68.82026-07-24
23GPT-5.5high93.2%65.32026-04-23
24GPT-5.6 Solxhigh93.1%66.62026-07-09
25GPT-6 Astralow93.1%65.92026-09-03
26Grok 4.5high93.1%59.32026-07-08
27MiniMax M392.9%57.72026-06-01
28GPT-5.6 Solhigh92.8%66.32026-07-09
29Qwen3.8 Maxmax92.8%63.22026-09-02
30Gemini 3.6 Flashhigh92.8%62.02026-07-21
31DeepSeek V4 Promax92.8%60.82026-08-13
32Claude Fable 5max92.6%69.32026-06-09
33GPT-5.6 Solmedium92.6%64.42026-07-09
34GPT-5.5medium92.6%62.92026-04-23
35GPT-5.6 Terramax92.5%64.42026-07-09
36Qwen3.7max92.3%60.62026-05-19
37Qwen3.8-Flash-Next92.3%60.32026-08-26
38Gemini 3.5 Flashhigh92.2%63.12026-05-19
39Gemini 3.7 Flashmedium92.1%63.12026-08-13
40Gemini 3.5 Flashmedium92.1%62.82026-05-19
41GPT-5.4xhigh92.0%64.72026-03-05
42Claude Opus 4.8max92.0%63.32026-05-28
43Gemini 3.8 Flashlow92.0%61.02026-09-02
44Claude Opus 5medium91.9%64.62026-07-24
45GLM 5.3max91.7%63.52026-08-18
46GPT-5.3-Codexxhigh91.5%63.82026-02-05
47Claude Opus 4.7max91.4%62.42026-04-16
48Deepseek v4 Flash Visionmax91.3%58.22026-08-21
49GLM 5.3 Flash91.2%62.32026-08-26
50GPT-5.6 Lunamax91.1%59.92026-07-09
51Kimi K2.6thinking91.1%59.42026-04-20
52Grok 4.20thinking91.1%58.62026-04-07
53Claude Sonnet 5max91.1%58.42026-06-30
54GPT-5.5low91.0%59.72026-04-23
55GPT-5.6 Terraxhigh90.8%62.82026-07-09
56Gemini 3 Prohigh90.8%60.02025-11-18
57DeepSeek V4 Flashmax90.8%59.42026-07-31
58Claude Fable 5.1high90.6%69.42026-09-01
59Agnes 2.5 Pro Beta90.5%59.12026-08-26
60DeepSeek V4 Pro 0424high90.5%57.82026-04-24

More reasoning benchmarks · All LLM benchmarks

Cite as: BenchLeader, “GPQA Diamond (AA) leaderboard”, data as of 26 Oct 2026, https://www.benchleader.com/benchmarks/aa_gpqa.

GPQA Diamond (AA): questions

What does GPQA Diamond (AA) measure?
GPQA Diamond is a set of 198 multiple-choice questions in biology, physics and chemistry written by PhD-level domain experts. The questions are designed to be "Google-proof": a skilled non-expert with unlimited web access still answers only about a third correctly, while experts in the field score around two-thirds. The Diamond subset keeps only the questions that experts answered correctly and non-experts got wrong. Scores are reported in percent of tasks solved; higher is better.
Which AI model leads GPQA Diamond (AA)?
GPT-6 Astra leads GPQA Diamond (AA) with 96.3% as of 26 Oct 2026, ahead of GPT-6 Astra at 96.1%.
How many models have GPQA Diamond (AA) results?
559 model configurations have a GPQA Diamond (AA) result on BenchLeader, all taken from Artificial Analysis.
Who runs GPQA Diamond (AA) and how often is it updated?
GPQA Diamond (AA) is published by Artificial Analysis. BenchLeader re-reads the published results once a day and records the date each result was published.
Does GPQA Diamond (AA) count toward the BenchLeader Index?
No. GPQA Diamond (AA) is shown for reference but left out of the composite index.