Home › Best › Largest context window
LLM with the largest context window
Ranked by published context window. A larger window is not free: long-context tiers raise the per-token rate above a threshold, and Anthropic charges flat across 1M. Check the long-context column before committing to a big prompt.
Top 15 by context window
| # | Model | Provider | Context | Input $/M | Output $/M | Long-context tier |
|---|---|---|---|---|---|---|
| 1 | Llama 4 Scout | Meta | 10,000,000 | $0.18 | $0.59 | none |
| 2 | DeepSeek V4.1 Flash | DeepSeek | 1,100,000 | $0.30 | $1.20 | none |
| 3 | GPT-6 Astra | OpenAI | 1,050,000 | $10 | $50 | above 272,000 tokens |
| 4 | GPT-6.1 Sol | OpenAI | 1,050,000 | $2 | $10 | above 272,000 tokens |
| 5 | GPT-6.1 Terra | OpenAI | 1,050,000 | $1 | $5 | above 272,000 tokens |
| 6 | GPT-6 Sol | OpenAI | 1,050,000 | $2 | $10 | above 272,000 tokens |
| 7 | GPT-6 Luna | OpenAI | 1,050,000 | $0.10 | $0.50 | above 272,000 tokens |
| 8 | GPT-5.6 Sol | OpenAI | 1,050,000 | $4 | $20 | above 272,000 tokens |
| 9 | GPT-5.6 Terra | OpenAI | 1,050,000 | $2 | $12 | above 272,000 tokens |
| 10 | GPT-5.6 Luna | OpenAI | 1,050,000 | $0.20 | $1.20 | above 272,000 tokens |
| 11 | GPT-5.5 | OpenAI | 1,050,000 | $5 | $30 | above 272,000 tokens |
| 12 | GPT-5.4 | OpenAI | 1,050,000 | $2.50 | $15 | above 272,000 tokens |
| 13 | GPT-4.1 | OpenAI | 1,047,576 | $2 | $8 | none |
| 14 | Claude Fable 5.1 | Anthropic | 1,000,000 | $10 | $50 | none |
| 15 | Claude Fable 5.2 | Anthropic | 1,000,000 | $8 | $40 | none |
Frequently asked
Which LLM has the largest context window?
Llama 4 Scout publishes a 10M-token window, the largest in this dataset, followed by DeepSeek V4.1 Flash at 1.1M and most current models at 1M tokens (GPT-6 family, Claude, Gemini, Grok 4.x).
Does a larger context window cost more per token?
Sometimes. OpenAI roughly doubles input above 272K tokens and Google raises Pro tiers above 200K. Anthropic and DeepSeek charge flat across the full window. The calculator applies these thresholds automatically.
Is 1M context actually usable?
Quality degrades well before the limit on most models, and a 1M-token prompt is expensive even at cheap rates. Measure retrieval quality at your real prompt size before sizing the window.
Browse all 45 models · Compare two models