Glossary
Context window
How much text the model can consider at once, adding up the prompt, the documents and the answer, in tokens, according to OpenRouter.
What it measures
How much text the model can consider at once, adding up the prompt, the documents and the answer, in tokens, according to OpenRouter.
When it matters
It matters for long contracts, codebases and chat history. A large window does not guarantee the model makes good use of everything that fits. If the file is long, I still check whether the answer cites the right part.
How to read it
The number is in tokens, according to OpenRouter. In this table, the filter cutoffs are 128 thousand, 256 thousand and 1 million. A large window can hold long contracts and codebases, but the model still has to find the right part.
Who leads
- Grok 4.20 xAI 2,000,000
- Grok 4.20 Multi-Agent xAI 2,000,000
- Llama 4 Scout Meta 1,310,720
- GPT-6.1 Sol OpenAI 1,050,000
- GPT-6 Astra OpenAI 1,050,000
- GPT-6 Luna OpenAI 1,050,000
- GPT-5.5 OpenAI 1,050,000
- MiMo V2.6 Pro Xiaomi 1,050,000
- GPT-5.4 OpenAI 1,050,000
- GPT-5.4 Pro OpenAI 1,050,000