Context window against price
Every model's window, the most text it holds at once, plotted against what it costs. Find the largest window for your money, and set the size of your document to see which models can read it whole.
Prices read from each maker's page on 14 October 2026
- Largest window
- 1.05M tokens on GPT-6 Astra, GPT-6.1 Sol, GPT-6 Luna
- Cheapest with 1M or more
- GPT-6 Luna, $0.10 in and $0.50 out
- The usual window now
- 1M tokens or more on 90% of the newest models
Window against price
Further right holds more text at once; higher costs more. The yellow squares, named on the chart, are the best value: none gives a window as large for less. Tap any square for its figures.
Not drawn: 3 models whose makers do not print a window on the price page.
about 225 pages
13 of 13 models here can read it whole. Cheapest to read once: GPT-6 Luna, 1.5 cents.
Best value, by window
The figures behind the chart
Largest window first. Prices are for 1 million tokens; the blend is our own figure, 3 tokens read for each 1 written.
Show the table
| Model | Window | Input | Output | Blend 3:1 |
|---|---|---|---|---|
| GPT-6 Luna · OpenAI | 1.05M | $0.10 | $0.50 | $0.20 |
| GPT-5.6 Luna · OpenAI | 1.05M | $0.20 | $1.20 | $0.45 |
| GPT-6.1 Sol · OpenAI | 1.05M | $2 | $10 | $4 |
| GPT-6 Sol · OpenAI | 1.05M | $2 | $10 | $4 |
| GPT-5.6 Terra · OpenAI | 1.05M | $2 | $12 | $4.50 |
| gpt-5.4 · OpenAI | 1.05M | $2.50 | $15 | $5.625 |
| GPT-5.6 Sol · OpenAI | 1.05M | $4 | $20 | $8 |
| gpt-5.5 · OpenAI | 1.05M | $5 | $30 | $11.25 |
| GPT-6 Astra · OpenAI | 1.05M | $10 | $50 | $20 |
| gpt-5.5-pro · OpenAI | 1.05M | $30 | $180 | $67.50 |
| gpt-5.4-pro · OpenAI | 1.05M | $30 | $180 | $67.50 |
| Gemini 2.5 Flash-Lite · Google | 1.05M | $0.10 | $0.40 | $0.175 |
| Gemini 3.1 Flash-Lite · Google | 1.05M | $0.25 | $1.50 | $0.5625 |
| Gemini 3.5 Flash-Lite · Google | 1.05M | $0.30 | $2.50 | $0.85 |
| Gemini 2.5 Flash · Google | 1.05M | $0.30 | $2.50 | $0.85 |
| Gemini 3 Flash Preview · Google | 1.05M | $0.50 | $3 | $1.125 |
| Gemini 3.8 Flash · Google | 1.05M | $0.75 | $3.75 | $1.50 |
| Gemini 3.6 Flash · Google | 1.05M | $0.75 | $3.75 | $1.50 |
| Gemini 2.5 Pro · Google | 1.05M | $1.25 | $10 | $3.4375 |
| Gemini 3.1 Pro Preview · Google | 1.05M | $2 | $12 | $4.50 |
| Kimi K3 · Moonshot AI | 1.05M | $3 | $15 | $6 |
| gpt-4.1-nano · OpenAI | 1.05M | $0.10 | $0.40 | $0.175 |
| gpt-4.1-mini · OpenAI | 1.05M | $0.40 | $1.60 | $0.70 |
| gpt-4.1 · OpenAI | 1.05M | $2 | $8 | $3.50 |
| Claude Haiku 5.5 · Anthropic | 1M | $0.10 | $0.50 | $0.20 |
| DeepSeek V4.1 Flash · DeepSeek | 1M | $0.30 | $1.20 | $0.525 |
| Grok 4.3 · xAI | 1M | $1.25 | $2.50 | $1.5625 |
| grok-4.20-0309-reasoning · xAI | 1M | $1.25 | $2.50 | $1.5625 |
| grok-4.20-0309-non-reasoning · xAI | 1M | $1.25 | $2.50 | $1.5625 |
| grok-4.20-multi-agent-0309 · xAI | 1M | $1.25 | $2.50 | $1.5625 |
| DeepSeek V4 Pro · DeepSeek | 1M | $1.32 | $3.96 | $1.98 |
| Claude Sonnet 5.5 · Anthropic | 1M | $2 | $10 | $4 |
| Claude Sonnet 5 · Anthropic | 1M | $2 | $10 | $4 |
| Claude Sonnet 4.6 · Anthropic | 1M | $3 | $15 | $6 |
| Claude Opus 5.5 · Anthropic | 1M | $4 | $20 | $8 |
| Claude Opus 5 · Anthropic | 1M | $5 | $25 | $10 |
| Claude Opus 4.8 · Anthropic | 1M | $5 | $25 | $10 |
| Claude Opus 4.7 · Anthropic | 1M | $5 | $25 | $10 |
| Claude Opus 4.6 · Anthropic | 1M | $5 | $25 | $10 |
| Claude Fable 5.1 · Anthropic | 1M | $10 | $50 | $20 |
| Claude Mythos 5.1 · Anthropic | 1M | $10 | $50 | $20 |
| Claude Fable 5 · Anthropic | 1M | $10 | $50 | $20 |
| Claude Mythos 5 · Anthropic | 1M | $10 | $50 | $20 |
| Grok 4.7 · xAI | 500K | $2 | $6 | $3 |
| Grok 4.6 · xAI | 500K | $2 | $6 | $3 |
| Grok 4.5 · xAI | 500K | $2 | $6 | $3 |
| gpt-5-nano · OpenAI | 400K | $0.05 | $0.40 | $0.1375 |
| gpt-5.4-nano · OpenAI | 400K | $0.20 | $1.25 | $0.4625 |
| gpt-5-mini · OpenAI | 400K | $0.25 | $2 | $0.6875 |
| gpt-5.4-mini · OpenAI | 400K | $0.75 | $4.50 | $1.6875 |
| gpt-5.1 · OpenAI | 400K | $1.25 | $10 | $3.4375 |
| gpt-5 · OpenAI | 400K | $1.25 | $10 | $3.4375 |
| gpt-5.2 · OpenAI | 400K | $1.75 | $14 | $4.8125 |
| gpt-5-pro · OpenAI | 400K | $15 | $120 | $41.25 |
| gpt-5.2-pro · OpenAI | 400K | $21 | $168 | $57.75 |
| Kimi K2.7 Code · Moonshot AI | 262K | $0.95 | $4 | $1.7125 |
| Kimi K2.6 · Moonshot AI | 262K | $0.95 | $4 | $1.7125 |
| Kimi K2.7 Code Highspeed · Moonshot AI | 262K | $1.90 | $8 | $3.425 |
| grok-build-0.1 · xAI | 256K | $1 | $2 | $1.25 |
| o4-mini · OpenAI | 200K | $1.10 | $4.40 | $1.925 |
| o3-mini · OpenAI | 200K | $1.10 | $4.40 | $1.925 |
| Claude Haiku 4.5 · Anthropic | 200K | $1 | $5 | $2 |
| o3 · OpenAI | 200K | $2 | $8 | $3.50 |
| Claude Sonnet 4.5 · Anthropic | 200K | $3 | $15 | $6 |
| Claude Opus 4.5 · Anthropic | 200K | $5 | $25 | $10 |
| o1 · OpenAI | 200K | $15 | $60 | $26.25 |
| o3-pro · OpenAI | 200K | $20 | $80 | $35 |
| o1-pro · OpenAI | 200K | $150 | $600 | $262.50 |
| gpt-4o-mini · OpenAI | 128K | $0.15 | $0.60 | $0.2625 |
| gpt-4o · OpenAI | 128K | $2.50 | $10 | $4.375 |
| gpt-3.5-turbo · OpenAI | 16K | $0.50 | $1.50 | $0.75 |
| gpt-3.5-turbo-instruct · OpenAI | 4K | $1.50 | $2 | $1.625 |
What the window holds, and what it costs you
The window is shared. Your instructions, every file you attach, the conversation so far and the answer the model writes all have to fit in it together. In a chat app the oldest turns quietly drop out once it is full; through the API the request is refused.
A large window is a ceiling, not a target. You pay for every token you send, on every request, so a 500-page document sent with each question costs 500 pages each time. A cache cuts the price of reading the same text again, and a search that pulls in only the pages that matter cuts the amount.
Models also find a detail less reliably when it sits deep in a very long prompt. Put what matters most near the start or the end, and test with your own documents before you count on the full window.
Questions
- What is a context window?
- The most tokens a model can hold at once: your instructions, the files you attach, the conversation so far and the answer it writes. Text past the window is not seen at all. A token is about three quarters of an English word.
- Which model has the largest context window?
- Among the newest models on our list, the largest window is 1.05M tokens, about 1,575 pages of text, on GPT-6 Astra, GPT-6.1 Sol, GPT-6 Luna.
- What is the cheapest model with a 1 million token window?
- GPT-6 Luna from OpenAI: $0.10 per million tokens read and $0.50 per million written, with a window of 1.05M.
- Is a bigger window always better?
- No. Every token you send is paid for on every request, models find details less reliably deep in a very long prompt, and some makers charge more once a prompt passes a size. Send what the task needs, and let a search pull in the rest.
- Why do some models cost more for long prompts?
- Reading a long prompt takes more computing than a short one, and some makers pass that on above a size such as 200,000 tokens. The chart plots the standard price; a model's details say where a higher one starts.