Context windows
What does a full context window actually cost?
SpaceXAI: Grok 4.20 has the largest tracked context window at 2.0M tokens - filling it costs $2.50 in input tokens alone. The cheapest way to fill a large context window right now is DeepSeek: DeepSeek V4 Flash 0423, at $0.09. Window size alone doesn't tell you the real cost of using it.
| # | Model | Context window | $/1M input | Cost to fill | Quality |
|---|---|---|---|---|---|
| 01 | SpaceXAI: Grok 4.20 | 2.0M | $1.25 | $2.50 | - |
| 02 | SpaceXAI: Grok 4.20 Multi-Agent | 2.0M | $1.25 | $2.50 | - |
| 03 | OpenAI: GPT-5.4 (batch) | 1.1M | $2.50 | $2.63 | 53.1 |
| 04 | OpenAI: GPT-5.4 Pro (batch) | 1.1M | $30.00 | $31.50 | - |
| 05 | OpenAI: GPT-5.5 (batch) | 1.1M | $5.00 | $5.25 | 56.3 |
| 06 | OpenAI: GPT-5.5 Pro (batch) | 1.1M | $30.00 | $31.50 | - |
| 07 | OpenAI: GPT-5.6 Luna (batch) | 1.1M | $0.10 | $0.10 | 52.3 |
| 08 | OpenAI: GPT-5.6 Luna Pro (batch) | 1.1M | $0.10 | $0.10 | - |
| 09 | OpenAI: GPT-5.6 Sol (batch) | 1.1M | $5.00 | $5.25 | 60.9 |
| 10 | OpenAI: GPT-5.6 Sol Pro (batch) | 1.1M | $5.00 | $5.25 | - |
| 11 | OpenAI: GPT-5.6 Terra (batch) | 1.1M | $1.00 | $1.05 | 56.6 |
| 12 | OpenAI: GPT-5.6 Terra Pro (batch) | 1.1M | $1.00 | $1.05 | - |
| 13 | Xiaomi: MiMo-V2.5 | 1.1M | $0.14 | $0.15 | 38 |
| 14 | Xiaomi: MiMo-V2.5-Pro | 1.1M | $0.43 | $0.46 | 42.9 |
| 15 | OpenAI GPT Latest | 1.1M | $5.00 | $5.25 | - |
| 16 | Meituan: LongCat 2.0 | 1.0M | $0.30 | $0.31 | - |
| 17 | DeepSeek: DeepSeek V4 Flash 0423 | 1.0M | $0.09 | $0.09 | 51.8 |
| 18 | DeepSeek: DeepSeek V4 Pro | 1.0M | $0.43 | $0.46 | 45.3 |
| 19 | Google: Gemini 2.5 Flash (batch) | 1.0M | $0.30 | $0.31 | - |
| 20 | Google: Gemini 2.5 Flash Lite (batch) | 1.0M | $0.10 | $0.10 | - |
| 21 | Google: Gemini 2.5 Pro (batch) | 1.0M | $1.25 | $1.31 | 25.9 |
| 22 | Google: Gemini 3 Flash Preview (batch) | 1.0M | $0.50 | $0.52 | - |
| 23 | Google: Gemini 3.1 Flash Lite (batch) | 1.0M | $0.25 | $0.26 | - |
| 24 | Google: Gemini 3.1 Flash Lite Preview | 1.0M | $0.25 | $0.26 | 25.6 |
| 25 | Google: Gemini 3.1 Pro Preview (batch) | 1.0M | $2.00 | $2.10 | 47.7 |
| 26 | Google: Gemini 3.1 Pro Preview Custom Tools | 1.0M | $2.00 | $2.10 | - |
| 27 | Google: Gemini 3.5 Flash (batch) | 1.0M | $1.50 | $1.57 | 52 |
| 28 | Google: Gemini 3.5 Flash Lite (batch) | 1.0M | $0.30 | $0.31 | 37.4 |
| 29 | Google: Gemini 3.6 Flash (batch) | 1.0M | $1.50 | $1.57 | 51.6 |
| 30 | Meta: Llama Guard 4 12B | 1.0M | $0.18 | $0.19 | - |
Cost to fill = context window (tokens) × prompt price per token. This is the cost of one request using the full window as input - real output tokens and multi-turn conversations cost more.
A big context window and a high-quality model aren't the same thing - among models with a window of 1M+ tokens, OpenAI: GPT-5.6 Sol (batch) scores highest on quality (see the Context window and Quality columns above).
Frequently asked questions
Which AI model has the largest context window?
SpaceXAI: Grok 4.20 has the largest tracked context window at 2.0M tokens.
How much does it cost to use a model's full context window?
Filling SpaceXAI: Grok 4.20's 2.0M-token context window costs $2.50 in input tokens at current pricing.