live
weekly refresh
basedagi.org
benchmark evidence

GPQA Diamond

Google-proof PhD-level science QA. 198 expert-crafted multiple-choice questions in biology, chemistry, physics.

winner on GPQA Diamond
direct benchmark result, not a broad vertical composite | source row dated 2000-01-01
scored on 2025-07-01 · stale source data (333d)
latest mapped results | top 20
#ModelScoreEvidenceTested
1Anthropic: Claude Opus 4.7
Anthropic
94.2
model-only
independent_benchmark
2000-01-01
2OpenAI: GPT-5.5
Openai
93.6
model-only
independent_benchmark
2000-01-01
3OpenAI: GPT-5.4
Openai
92.8
model-only
independent_benchmark
2000-01-01
4OpenAI: GPT-5.2
Openai
92.4
model-only
independent_benchmark
2000-01-01
5Anthropic: Claude Opus 4.6
Anthropic
91.3
model-only
independent_benchmark
2000-01-01
6DeepSeek: DeepSeek V4 Pro
Deepseek
90.1
model-only
independent_benchmark
2000-01-01
7Anthropic: Claude Sonnet 4.6
Anthropic
89.9
model-only
independent_benchmark
2000-01-01
8OpenAI: GPT-5.1
Openai
88.1
model-only
independent_benchmark
2000-01-01
9DeepSeek: DeepSeek V4 Flash
Deepseek
88.1
model-only
independent_benchmark
2000-01-01
10OpenAI: GPT-5.4 Mini
Openai
88.0
model-only
independent_benchmark
2000-01-01
11Anthropic: Claude Opus 4.5
Anthropic
87.0
model-only
independent_benchmark
2000-01-01
12MoonshotAI: Kimi K2 Thinking
Moonshotai
84.5
model-only
independent_benchmark
2000-01-01
13OpenAI: o3 Mini
Openai
77.2
model-only
independent_benchmark
2025-07-01
14OpenAI: o1
Openai
73.3
model-only
independent_benchmark
2025-07-01
15OpenAI: GPT-4.1
Openai
66.3
model-only
independent_benchmark
2025-07-01
16OpenAI: GPT-5
Openai
46.0
model-only
independent_benchmark
2025-07-01
what this result means

Google-proof PhD-level science QA. 198 expert-crafted multiple-choice questions in biology, chemistry, physics.

This benchmark contributes direct public evidence. Read its scope before generalizing the result.

A win here is a win on GPQA Diamond. Broad task pages require independent corroboration before naming a general winner.

source record
category: reasoning
metric: accuracy
matched models: 16
latest source date: 2025-07-01
direction: higher is better
inspect upstream source ->