Glossary
GPQA Diamond (OpenRouter)
This is the same GPQA Diamond test, run by OpenRouter Benchmarks (CC BY 4.0, latest version we checked) on the provider it routes that call to.
What it measures
This is the same GPQA Diamond test, run by OpenRouter Benchmarks (CC BY 4.0, latest version we checked) on the provider it routes that call to.
When it matters
It helps you see the model the way you are likely to call it through an API. The score can differ from Epoch AI because of the provider, the settings and the number of runs. If both appear on the page, I read both and I do not mix them in the same comparison.
How to read it
Compare it with the Epoch AI score for the same model. Both measure the same test, but they can differ because of the provider, the settings and the number of runs. I read both and I do not mix them in the same comparison.
Who leads
- Gemini 4 Argon Google96.5%
- Gemini 3.8 Flash Google95.6%
- GPT-6 Astra Pro OpenAI95.5%
- Gemini 3.1 Pro Google95.2%
- GPT-6.1 Sol OpenAI94.8%
- Ling 3.1 Flash Ant Group94.8%
- GPT-6 Astra OpenAI94.7%
- Gemini 3.7 Flash Google94.3%
- GPT-5.6 Sol Pro OpenAI94.1%
- GPT-5.5 OpenAI93.9%
Source: OpenRouter Benchmarks, latest version we checked.