One dimension is not a complete decision
Cost, token volume, capacity, and evidence coverage answer different questions. Compare them with a representative prompt, actual output length, retry behavior, and your quality bar.
Find a model for a taskSource-tracked capacity
AI model context window rankings
Compare documented context capacity and maximum output limits across tracked models. A larger window may help a workload, but it does not prove better answers or lower cost.
Source-tracked capacity
Largest documented context windows
Context and output limits come from the model records and their specification sources. Verify endpoint, tier, and effective limits before production use.
| Rank | Model | Provider | Context window | Max output | Specs checked |
|---|---|---|---|---|---|
| 1 | Gemini 1.5 Pro | 2,097,152 | 8,192 | 2026-07-17 | |
| 2 | GPT-5.6 Luna | OpenAI | 1,050,000 | 128,000 | 2026-07-13 |
| 3 | GPT-5.6 Sol | OpenAI | 1,050,000 | 128,000 | 2026-07-13 |
| 4 | GPT-5.6 Terra | OpenAI | 1,050,000 | 128,000 | 2026-07-13 |
| 5 | Gemini 1.5 Flash | 1,048,576 | 8,192 | 2026-07-17 | |
| 6 | Gemini 2.5 Flash | 1,048,576 | 65,536 | 2026-07-13 | |
| 7 | Gemini 2.5 Flash-Lite | 1,048,576 | 65,536 | 2026-07-13 | |
| 8 | Gemini 2.5 Pro | 1,048,576 | 65,536 | 2026-07-13 | |
| 9 | Gemini 3 Flash Preview | 1,048,576 | 65,536 | 2026-07-14 | |
| 10 | Gemini 3.1 Flash-Lite | 1,048,576 | 65,536 | 2026-07-14 | |
| 11 | Gemini 3.1 Pro Preview | 1,048,576 | 65,536 | 2026-07-14 | |
| 12 | Gemini 3.5 Flash | 1,048,576 | 65,536 | 2026-07-14 | |
| 13 | Gemini 3.5 Flash-Lite | 1,048,576 | 65,536 | 2026-07-21 | |
| 14 | Gemini 3.6 Flash | 1,048,576 | 65,536 | 2026-07-21 | |
| 15 | Kimi K3 | Moonshot AI | 1,048,576 | Requires verification | 2026-07-17 |
| 16 | Claude Fable 5 | Anthropic | 1,000,000 | 128,000 | 2026-07-14 |
| 17 | Claude Mythos 5 | Anthropic | 1,000,000 | 128,000 | 2026-07-14 |
| 18 | Claude Opus 4.8 | Anthropic | 1,000,000 | 128,000 | 2026-07-13 |
| 19 | Claude Opus 5 | Anthropic | 1,000,000 | 128,000 | 2026-07-28 |
| 20 | Claude Sonnet 4.6 | Anthropic | 1,000,000 | 128,000 | 2026-07-13 |
| 21 | Claude Sonnet 5 | Anthropic | 1,000,000 | 128,000 | 2026-07-13 |
| 22 | DeepSeek V4 Flash non-thinking | DeepSeek | 1,000,000 | 384,000 | 2026-07-13 |
| 23 | DeepSeek V4 Pro | DeepSeek | 1,000,000 | 384,000 | 2026-07-13 |
| 24 | Grok 4.5 | xAI | 500,000 | Requires verification | 2026-07-17 |
| 25 | gpt-5.4-mini | OpenAI | 400,000 | 128,000 | 2026-07-13 |
| 26 | gpt-5.4-nano | OpenAI | 400,000 | 128,000 | 2026-07-13 |
| 27 | Kimi K2.6 | Moonshot AI | 262,144 | Requires verification | 2026-07-17 |
| 28 | Kimi K2.7 Code | Moonshot AI | 262,144 | Requires verification | 2026-07-17 |
| 29 | Kimi K2.7 Code High-Speed | Moonshot AI | 262,144 | Requires verification | 2026-07-17 |
| 30 | Command A | Cohere | 256,000 | 8,000 | 2026-07-17 |
| 31 | Mistral Medium 3.5 | Mistral | 256,000 | Requires verification | 2026-07-17 |
| 32 | Claude Haiku 4.5 | Anthropic | 200,000 | 64,000 | 2026-07-13 |
| 33 | Gemini 3.5 Live Translate Preview | 131,072 | 65,536 | 2026-07-14 | |
| 34 | GPT-OSS 120B on Groq | Groq | 131,072 | 65,536 | 2026-07-15 |
| 35 | GPT-OSS 20B on Groq | Groq | 131,072 | 65,536 | 2026-07-15 |
| 36 | Llama 4 Scout on Groq | Groq | 131,072 | 8,192 | 2026-07-15 |
| 37 | Qwen3-32B on Groq | Groq | 131,072 | 40,960 | 2026-07-15 |
| 38 | Command R 08-2024 | Cohere | 128,000 | 4,000 | 2026-07-17 |
| 39 | Command R+ 08-2024 | Cohere | 128,000 | 4,000 | 2026-07-17 |
| 40 | GPT-4o | OpenAI | 128,000 | 16,384 | 2026-07-13 |
| 41 | GPT-4o mini | OpenAI | 128,000 | 16,384 | 2026-07-13 |
A context window is a documented capacity limit, not a guarantee that a task will use the full window well. Check the context window calculator before choosing a model.
Turn a ranking into a test
FAQ
Ranking methodology questions
Does a higher ranking mean a better model?
No. Each board measures one defined dimension. Token count, price, context capacity, and evidence coverage do not substitute for testing your own completed tasks.