Home / AI context window comparison
AI context window comparison
The context window is how much text a model can take in at once. Most current models now accept about 1 million tokens, but four charge more once a prompt passes a set length.
Prices checked
Quick answer
OpenAI’s four GPT-6 models have the largest context windows we track, at 1.05 million tokens, with Google’s three Gemini 4 models just behind. Thirteen more models accept 1 million tokens. Grok 4.7 has 500K, and Mistral Medium 3.5, Doubao Seed 2.0 Pro and Tencent Hy3 have 256K. Xiaomi doesn’t state a context window for MiMo-v2.6 Pro.
Every model ranked by context window
API context windows in tokens, largest first. Word and page counts are rough estimates for English text, at 0.75 words per token and 500 words per page.
| Model | Context window | About, in words | About, in pages | Long-prompt pricing |
|---|---|---|---|---|
| GPT-6.2 Astra OpenAI | 1.05M | 787,500 | 1,575 | None listed |
| GPT-6.2 Sol OpenAI | 1.05M | 787,500 | 1,575 | None listed |
| GPT-6.1 Terra OpenAI | 1.05M | 787,500 | 1,575 | None listed |
| GPT-6 Luna OpenAI | 1.05M | 787,500 | 1,575 | None listed |
| Gemini 4 Carbon Google | 1.05M | 786,000 | 1,572 | None listed |
| Gemini 4 Argon Google | 1.05M | 786,000 | 1,572 | None listed |
| Gemini 4 Hydrogen Google | 1.05M | 786,000 | 1,572 | Flat price across the window |
| Claude Fable 5.2 Anthropic | 1M | 750,000 | 1,500 | None listed |
| Claude Opus 5.5 Anthropic | 1M | 750,000 | 1,500 | None listed |
| Claude Sonnet 5.5 Anthropic | 1M | 750,000 | 1,500 | None listed |
| Claude Haiku 5.5 Anthropic | 1M | 750,000 | 1,500 | Higher rate above 100K tokens |
| Muse Wonder Meta | 1M | 750,000 | 1,500 | None listed |
| Muse Spark 1.3 Meta | 1M | 750,000 | 1,500 | None listed |
| DeepSeek V4 Pro DeepSeek | 1M | 750,000 | 1,500 | None listed |
| DeepSeek V4.1 Flash DeepSeek | 1M | 750,000 | 1,500 | None listed |
| Mistral Large 4 Mistral | 1M | 750,000 | 1,500 | None listed |
| Qwen4 Max Alibaba | 1M | 750,000 | 1,500 | Single price tier |
| Kimi K3.1 Pro Moonshot | 1M | 750,000 | 1,500 | None listed |
| GLM-5.3 Zhipu | 1M | 750,000 | 1,500 | None listed |
| MiniMax M3 MiniMax | 1M | 750,000 | 1,500 | Higher rate above 512K tokens |
| Grok 4.7 xAI | 500K | 375,000 | 750 | Higher rate from 200K tokens |
| Mistral Medium 3.5 Mistral | 256K | 192,000 | 384 | None listed |
| Doubao Seed 2.0 Pro ByteDance | 256K | 192,000 | 384 | Higher rate above 128K tokens |
| Hy3 Tencent | 256K | 192,000 | 384 | None listed |
| MiMo-v2.6 Pro Xiaomi | Not stated | Not stated | Not stated | None listed |
What a context window means
- It’s measured in tokens. A token is a chunk of text, often part of a word. A common rule of thumb for English is that 1 token is about three quarters of a word.
- It covers everything in the request. Your instructions, any files or chat history you send, and the model’s reply all share the same window.
- 1 million tokens is a lot. It’s roughly 750,000 words, or about 1,500 pages at 500 words a page. That’s several long novels in one prompt.
- 256K is still large. It’s about 192,000 words, or close to 400 pages.
- Other languages and code differ. Text in Chinese, Japanese or source code uses tokens at a different rate, so treat the word counts as estimates.
- A bigger window isn’t a quality score. It tells you how much fits, not how well the model uses it.
Where long prompts cost more
Four models we track raise their price when a prompt passes a set length. The rates below are input / output per million tokens.
| Model | Standard price | Threshold | Price above the threshold |
|---|---|---|---|
| Claude Haiku 5.5 | $0.10 / $0.50 | Prompts over 100K tokens | $0.50 / $2.50 |
| Grok 4.7 | $2 / $6 | Prompts of 200K tokens or more | $4 / $12, billed on every token in the request |
| Doubao Seed 2.0 Pro | $0.50 / $3 | Prompts over 128K tokens | $1 / $6 |
| MiniMax M3 | $0.30 / $1.20 | Prompts over 512K tokens | $0.60 / $2.40 |
Flat pricing across the full window
Some companies state that the price doesn’t change with prompt length.
- Gemini 4 Hydrogen: Google says it’s priced flat across the full 1 million token window, at $1.50 / $7.50 per million tokens.
- Claude Fable, Opus 5.5 and Sonnet 5.5: Anthropic says its current models, other than Haiku 5.5, include the full 1 million token window at standard pricing.
- Qwen4 Max: Alibaba lists a single price tier up to 1 million tokens, at $1.50 / $4.50 per million tokens.
- OpenAI: the API pricing page shows separate short and long context settings for its GPT-6 models. We couldn’t confirm the long context rates, so they aren’t listed here.
- A full prompt still costs real money. One 1 million token prompt to Claude Fable 5.2 costs $8 in input alone. Try your own numbers in the AI cost calculator.
Context in chat apps is different
The numbers above are for developers using the API. Chat apps often allow less, and the limit can depend on your plan.
- ChatGPT: the plan page lists a 27K token window for instant replies on Free, 54K on Go and Plus, and 128K on Pro. Reasoning models get 256K on Go and Plus and 400K on Pro.
- Gemini: Google lists a 1 million token context window as a feature of Google AI Pro.
- Other apps: most don’t publish a per-plan context figure, so we don’t list one.
- If you mainly need to load long documents, compare plans in ChatGPT vs Gemini or model prices in cheapest AI API.
Questions people ask
Which AI model has the largest context window?
Among the 25 models we track, OpenAI’s GPT-6.2 Astra, GPT-6.2 Sol, GPT-6.1 Terra and GPT-6 Luna are largest at 1.05 million tokens. Google’s Gemini 4 models follow at 1,048K tokens, which also rounds to 1.05 million.
How many words is 1 million tokens?
Roughly 750,000 English words, or about 1,500 pages at 500 words a page. That’s an estimate: the exact count depends on the text and the model’s tokenizer.
Does a longer prompt cost more per token?
On most models, no. Four models we track charge a higher rate for long prompts: Claude Haiku 5.5 above 100K tokens, Grok 4.7 from 200K, Doubao Seed 2.0 Pro above 128K and MiniMax M3 above 512K.
What is the context window of MiMo-v2.6 Pro?
Xiaomi’s pricing page doesn’t state one, so we list it as not stated. See Chinese AI models for its price.
Sources: Claude API pricing (Haiku 5.5 long-prompt rates), xAI API pricing (Grok 4.7 long context rates), BytePlus model pricing (Doubao Seed 2.0 Pro), MiniMax pricing, Alibaba Cloud Model Studio pricing (Qwen4 Max), Google: introducing Gemini 4 Hydrogen, ChatGPT plans, Google AI plans, OpenAI API pricing, OpenAI: what are tokens and how to count them. Checked November 18, 2026.