Benchmarks / Knowledge & Science

GPQA Diamond

Graduate-level, Google-proof science questions (diamond subset).

maintainer unknown · 631 observations on record

Versions

1 version

Version	Tasks	Window	Released	Notes
GPQA Diamond current gpqa-diamond	198	—	—

Current leaderboard

GPQA Diamond · Accuracy · top 25 of 547 systems

#	System	Accuracy %	Verification	Source
1	Gemini 3.1 Pro Previewhigh effort Google DeepMind	95.5%	independent	Vals AI
2	GPT 5.6 Solmax effort OpenAI	95.2%	independent	Vals AI
3	GPT 5.4 Pro (2026-03-05)xhigh effort OpenAI	94.6%	independent	Epoch AI Benchmarking Hub 2026-03-20
4	Gemini 3.1 Pro Previewhigh effort Google DeepMind	94.4%	independent	Epoch AI Benchmarking Hub 2026-08-06
5	Gemini 3.1 Pro Preview Google DeepMind	94.1%	independent	Artificial Analysis
6	GPT 5.6 Solmax effort OpenAI	94.1%	independent	Artificial Analysis
7	GPT 5.5 Pre Releasexhigh effort OpenAI	94.0%	independent	Epoch AI Benchmarking Hub 2026-04-24
8	GPT 5.5 Pro Pre Releasexhigh effort OpenAI	93.9%	independent	Epoch AI Benchmarking Hub 2026-04-24
9	GPT 5.5 (2026-04-22)xhigh effort OpenAI	93.5%	independent	Artificial Analysis
10	Grok 4.5high effort xAI	93.4%	independent	Epoch AI Benchmarking Hub 2026-07-08
11	GPT 5.6 Terramax effort OpenAI	93.3%	independent	Epoch AI Benchmarking Hub 2026-07-09
12	GPT 5.4 (2026-03-05)xhigh effort OpenAI	93.3%	independent	Epoch AI Benchmarking Hub 2026-03-06
13	GPT 5.5 (2026-04-22)high effort OpenAI	93.2%	independent	Artificial Analysis
14	Claude Fable 5max effort Anthropic	93.2%	independent	Vals AI
15	GPT 5.5 (2026-04-22)xhigh effort OpenAI	93.2%	independent	Vals AI
16	GPT 5.6 Solxhigh effort OpenAI	93.1%	independent	Artificial Analysis
17	Grok 4.5high effort xAI	92.9%	independent	Vals AI
18	Minimax M3 MiniMax	92.9%	independent	Artificial Analysis
19	Gemini 3.5 Flashhigh effort Google DeepMind	92.8%	independent	Epoch AI Benchmarking Hub 2026-05-22
20	GPT 5.6 Solhigh effort OpenAI	92.8%	independent	Artificial Analysis
21	Gemini 3.5 Flashhigh effort Google DeepMind	92.7%	independent	Vals AI
22	Minimax M3 MiniMax	92.7%	independent	Vals AI
23	Gemini 3 Pro Preview Google DeepMind	92.6%	independent	Epoch AI Benchmarking Hub 2025-11-19
24	GPT 5.6 Solmedium effort OpenAI	92.6%	independent	Artificial Analysis
25	GPT 5.5 (2026-04-22)medium effort OpenAI	92.6%	independent	Artificial Analysis

One row per evaluated system — reasoning-effort variants rank separately. Where a system has attr variants for this version (eval splits, style control), only the canonical variant is ranked: split=semi_private, style_control=true, or the unsplit run.

Source agreement

where evaluating organizations agree — and don't

GPQA DiamondArtificial Analysis vs Epoch AI Benchmarking Hub

overlap 61 systemsrank correlation 0.96mean |Δ| 2.56max |Δ| 26.18

most contested: O1 Preview (2024-09-12) (+26.18) · Magistral Small 2509 (+18.7) · Deepseek R1 Distill Llama 70B (-15.54)

Possible reasons: run setting 'implementation' differs (artificial-analysis vs —).

GPQA DiamondArtificial Analysis vs Vals AI

overlap 61 systemsrank correlation 0.95mean |Δ| 3.17max |Δ| 39.95

most contested: Mistral Medium 3.5 (+39.95) · Mimo V2 Flash (+25.26) · Magistral Medium 2509 (+11.53)

Possible reasons: run setting 'implementation' differs (artificial-analysis vs vals-ai).

GPQA DiamondEpoch AI Benchmarking Hub vs Vals AI

overlap 68 systemsrank correlation 0.95mean |Δ| 2.51max |Δ| 10.73

most contested: Magistral Small 2509 (-10.73) · GPT OSS 20B (-8.14) · Claude Fable 5 (-7.32)

Possible reasons: run setting 'implementation' differs (— vs vals-ai).