LLM Context Window & Capacity Benchmarks
Context window benchmarks test the absolute payload threshold of large language models. Llama 4 Scout holds the highest context capacity at 10,000,000 tokens (~7.5M words), followed by Gemini 3.5 Flash and Claude Sonnet 4.6 at 1,048,576 tokens (~780K words), while GPT-4o sits at 128,000 tokens (~96K words).
Logical Context Capacity Rankings
Comparison of maximum document capacities (books, pages, codebases) mapped to absolute token context windows.
| Model Name | Tokenizer Provider | Context Window (Tokens) | Max Output Window | Equivalent Page Capacity | Detailed Guide |
|---|---|---|---|---|---|
| Llama 4 Scout (17B) | huggingface | 10,000,000 | 8,192 tokens | ~15,000 pages | View details → |
| GPT-5.5 | openai | 1,050,000 | 128,000 tokens | ~1,575 pages | View details → |
| GPT-5.5 Pro | openai | 1,050,000 | 128,000 tokens | ~1,575 pages | View details → |
| Gemini 3.5 Flash | gemini | 1,048,576 | 65,536 tokens | ~1,573 pages | View details → |
| Gemini 3.1 Pro | gemini | 1,048,576 | 65,536 tokens | ~1,573 pages | View details → |
| Gemini 3 Flash | gemini | 1,048,576 | 65,536 tokens | ~1,573 pages | View details → |
| Gemini 3.1 Flash-Lite | gemini | 1,048,576 | 65,536 tokens | ~1,573 pages | View details → |
| Gemini 2.5 Pro | gemini | 1,048,576 | 65,536 tokens | ~1,573 pages | View details → |
| Gemini 2.5 Flash | gemini | 1,048,576 | 65,535 tokens | ~1,573 pages | View details → |
| Gemini 2.5 Flash-Lite | gemini | 1,048,576 | 8,192 tokens | ~1,573 pages | View details → |
| Claude Opus 4.8 | anthropic | 1,000,000 | 128,000 tokens | ~1,500 pages | View details → |
| Claude Fable 5 | anthropic | 1,000,000 | 128,000 tokens | ~1,500 pages | View details → |
| Claude Opus 4.7 | anthropic | 1,000,000 | 128,000 tokens | ~1,500 pages | View details → |
| Claude Opus 4.6 | anthropic | 1,000,000 | 128,000 tokens | ~1,500 pages | View details → |
| Claude Sonnet 4.6 | anthropic | 1,000,000 | 64,000 tokens | ~1,500 pages | View details → |
| Llama 4 Maverick (17B) | huggingface | 1,000,000 | 8,192 tokens | ~1,500 pages | View details → |
| DeepSeek V4 Flash | huggingface | 1,000,000 | 8,192 tokens | ~1,500 pages | View details → |
| DeepSeek V4 Pro | huggingface | 1,000,000 | 384,000 tokens | ~1,500 pages | View details → |
| GPT-5.4 mini | openai | 400,000 | 128,000 tokens | ~600 pages | View details → |
| GPT-5.4 nano | openai | 272,000 | 128,000 tokens | ~408 pages | View details → |
| Qwen 3.5 9B | huggingface | 262,144 | 4,096 tokens | ~393 pages | View details → |
| Qwen 3.6 27B | huggingface | 262,144 | 65,536 tokens | ~393 pages | View details → |
| Mistral Small 4 (119B) | huggingface | 256,000 | 8,192 tokens | ~384 pages | View details → |
| Kimi K2.6 | huggingface | 256,000 | 8,192 tokens | ~384 pages | View details → |
| Gemma 4 26B A4B | huggingface | 256,000 | 8,192 tokens | ~384 pages | View details → |
| Gemma 4 31B | huggingface | 256,000 | 262,000 tokens | ~384 pages | View details → |
| o3 | openai | 200,000 | 100,000 tokens | ~300 pages | View details → |
| o4-mini | openai | 200,000 | 100,000 tokens | ~300 pages | View details → |
| Claude Opus 4.5 | anthropic | 200,000 | 8,000 tokens | ~300 pages | View details → |
| Claude Sonnet 4.5 | anthropic | 200,000 | 64,000 tokens | ~300 pages | View details → |
| Claude Haiku 4.5 | anthropic | 200,000 | 64,000 tokens | ~300 pages | View details → |
| GPT-4o | openai | 128,000 | 16,384 tokens | ~192 pages | View details → |
| GPT-4o mini | openai | 128,000 | 16,384 tokens | ~192 pages | View details → |
| Gemma 4 E4B | huggingface | 128,000 | 4,096 tokens | ~192 pages | View details → |
| GPT-3.5 Turbo | openai | 16,385 | 4,096 tokens | ~25 pages | View details → |
| GPT-4 | openai | 8,192 | 4,096 tokens | ~12 pages | View details → |