GPQA Diamond
GPQA Diamond - Google-Proof Q&A benchmark for graduate-level scientific reasoning.
Category
STEM & Reasoning
Score type
percent
Scale
100%
higher is better
Models tested
57
What is GPQA Diamond?
GPQA Diamond is the 'diamond' subset of the Google-Proof Q&A benchmark — the hardest and cleanest part. The questions are written and checked by PhD scientists and are designed so that a skilled non-expert with full internet access can still barely get them right. The model has to answer graduate-level science questions with four multiple-choice options.
What does it test?
GPQA Diamond measures hard reasoning in science, technology, engineering and math. In detail, it checks:
- expert-level physics, chemistry and biology knowledge
- real reasoning instead of memorised question patterns
- resistance to plausible distractors
- reliability without web access (the questions are not Googleable)
How to read the score
The score is accuracy: the percentage of tasks answered or solved correctly. 100% means every task was solved perfectly; higher is better.
Frequently asked questions
What is the GPQA Diamond benchmark?
GPQA Diamond is the 'diamond' subset of the Google-Proof Q&A benchmark — the hardest and cleanest part. The questions are written and checked by PhD scientists and are designed so that a skilled non-expert with full internet access can still barely get them right. The model has to answer graduate-level science questions with four multiple-choice options.
What does GPQA Diamond test?
GPQA Diamond mainly measures hard reasoning in science, technology, engineering and math. Concretely, it checks: expert-level physics, chemistry and biology knowledge; real reasoning instead of memorised question patterns; resistance to plausible distractors.
How is the GPQA Diamond score calculated?
The score is accuracy: the percentage of tasks answered or solved correctly. 100% means every task was solved perfectly; higher is better. The declared maximum score of this benchmark is 100.0.
Where do I get GPQA Diamond data and scores?
The official source is https://huggingface.co/datasets/Idavidrein/gpqa. The official benchmark page/reference is linked on this page. Every score on Open LLM Radar is also available as open data: download the scores JSON (http://open-llm-radar-api:8000/api/v1/benchmarks/gpqa-diamond/scores) or pull it directly from the Open LLM Radar API with any HTTP client.
Benchmark data & downloads
Where to get, inspect and reuse the data behind this benchmark.
Official benchmark page / paper
The benchmark maintainer's website, paper or leaderboard
Benchmark dataset
The question/task data itself — download and evaluate locally
Benchmark metadata (JSON)
This page's benchmark record from the Open LLM Radar API
All model scores (JSON download)
Full score table as JSON; paginate with limit/offset params
Open LLM Radar API docs
Swagger UI to explore and reuse all radar API endpoints
Model scores 57 models tested
| # | Model | Score | Date | Source | Verified |
|---|---|---|---|---|---|
| 26 |
Ornith-1.0-35B-A3B
Ornith AI
|
86.2%
|
19.08.2026 | Report | Unverified |
| 27 |
Qwen3.6-35B-A3B
Qwen Team (Alibaba Cloud)
|
86%
|
16.04.2026 | Qwen Team (official model card); KRAFTON re-eval (Raon-1-mini-preview card, gpt-5.6-luna harness): 83.2 | Verified |
| 28 |
Qwen3.5-27B
Qwen Team (Alibaba Cloud)
|
85.5%
|
16.04.2026 | Qwen Team (official model card - comparison) | Unverified |
| 29 |
Ripple-65B Preview
Ricursive Intelligence
|
85.2%
|
— | Report | Verified |
| 30 |
Gemma4-31B-it
Google
|
84.3%
|
16.04.2026 | Qwen Team (official model card - comparison) | Unverified |
| 31 |
Gemma-4-31B
Google
|
84.3%
|
19.08.2026 | Report | Unverified |
| 32 |
Kolibri-1
Aleph Alpha
|
84.3%
|
03.10.2026 | Report | Verified |
| 33 |
Raon-1 mini
KRAFTON AI
|
84.3%
|
21.10.2026 | Report | Verified |
| 34 |
Qwen3.5-35B-A3B
Qwen Team (Alibaba Cloud)
|
84.2%
|
16.04.2026 | Qwen Team (official model card - comparison) | Unverified |
| 35 |
Muse-Glimmer-30B
Meta Superintelligence Lab
|
83.5%
|
01.08.2026 | Meta Superintelligence Lab (official model card); KRAFTON re-eval (Raon-1-mini-preview card, gpt-5.6-luna harness): 82.6 | Verified |
| 36 |
Gemma4-26B-A4B
Google
|
82.3%
|
16.04.2026 | Qwen Team (official model card - comparison); KRAFTON re-eval (Raon-1-mini-preview card, gpt-5.6-luna harness): 80.7 | Unverified |
| 37 |
Qwen3-235B-A22B
Qwen Team (Alibaba Cloud)
|
81.1%
|
01.02.2026 | Report | Unverified |
| 38 |
GPT-OSS-120B
OpenAI
|
80.1%
|
08.08.2025 | Report | Verified |
| 39 |
Gemma4-12B
Google
|
78.8%
|
02.07.2026 | Report | Verified |
| 40 |
Qwen3.5-4B
Qwen
|
76.2%
|
01.02.2026 | Report | Verified |
| 41 |
LFM2.5-24B-A2B
LiquidAI
|
74.42%
|
13.10.2026 | Report | Verified |
| 42 |
Qwen3-30B-A3B-Thinking-2507
Qwen
|
73.4%
|
31.07.2025 | Report | Verified |
| 43 |
GPT-OSS-20B
OpenAI
|
71.5%
|
05.08.2025 | Report | Verified |
| 44 |
MiniCPM5-2B
OpenBMB
|
70.2%
|
— | Report | Verified |
| 45 |
Qwen3-30B-A3B
Qwen
|
65.8%
|
29.04.2025 | Report | Verified |
| 46 |
LFM3-1.7B
LiquidAI
|
63.37%
|
— | Report | Verified |
| 47 |
Gemma4-E4B
Google
|
58.6%
|
02.07.2026 | Report | Verified |
| 48 |
granite-4.2-3B
IBM
|
55.9%
|
— | Report | Verified |
| 49 |
LFM2.5-2.6B
LiquidAI
|
55.8%
|
— | Report | Verified |
| 50 |
Nemotron-3-Nano-4B
NVIDIA
|
51.3%
|
— | Report | Verified |