Skip to content

PhD-level science questions (GPQA Diamond)

Best independently measured score, by model release date. Multiple choice, 4 options.

Topic: Benchmarks

What this measures: 198 hard biology, chemistry and physics questions written by PhDs. Experts in the field get about 70% right; smart non-experts with Google get 34%; guessing gets 25%.

  • Best AI score (Epoch AI runs)
  • Reference level
20%40%60%80%100%202420252026PhD experts in the field (69.7%)Skilled non-experts with web access (34%)

Saturated: Top scores (about 96%) are near the test's ceiling, where errors in the answer key cap what's achievable, so differences between today's best models are mostly noise. Now watch: Humanity's Last Exam.

Many runs happened after the model's release date. The expert baseline comes from a study OpenAI ran with recruited PhDs; the original paper reports 65% on the full question set. The questions have been public since 2023, so models may have seen them in training.

Sources: Epoch AI, Rein et al.Data as of 9 Nov 2026
Show data table and sources
DateLabelValue
14 Mar 2023GPT-4 (0314)36%
6 Nov 2023GPT-4 Turbo42%
29 Feb 2024Claude 3 Opus47%
13 May 2024GPT-4o49%
20 Jun 2024Claude 3.5 Sonnet54%
12 Sept 2024o1-mini (high)62%
17 Dec 2024o1 (high)77%
31 Jan 2025o3-mini (high)77%
24 Feb 2025Claude 3.7 Sonnet (thinking)79%
25 Mar 2025Gemini 2.5 Pro84%
5 Jun 2025Gemini 2.5 Pro (06-05)85%
9 Jul 2025Grok 487%
13 Nov 2025GPT-5.1 (high)88%
18 Nov 2025Gemini 3 Pro93%
19 Feb 2026Gemini 3.1 Pro (high)94%
5 Mar 2026GPT-5.4 Pro95%
13 Aug 2026Gemini 3.7 Flash (high)95%
2 Sept 2026Gemini 3.8 Flash (high)95%
3 Sept 2026GPT-6 Astra (max)96%
30 Sept 2026Gemini 4 Argon (high)96%
PhD experts in the field (69.7%)70%
Skilled non-experts with web access (34%)34%

Entries with this chart