Anthropic · 200K context · updated 2026-10-02
Input · what you send
$1
per 1M tokens
Output · what it writes
$5
per 1M tokens
2,000,000 tokens
10,000,000 tokens
For context: the cheapest model tracked here is GPT-6 Luna at $0.50/1M output tokens, and the most expensive is GPT-6 Astra at $50 — a 100× spread.Try your own budget in the calculator →
| Model | Input / 1M | Output / 1M | vs Claude Haiku 4.5 (output) |
|---|---|---|---|
| GPT-6 Astra | $10 | $50 | 10× pricier |
| Claude Fable 5.1 | $10 | $50 | 10× pricier |
| Claude Opus 5 | $5 | $25 | 5× pricier |
| Claude Opus 5.5 | $4 | $20 | 4× pricier |
| GPT-6 Sol | $2 | $10 | 2× pricier |
| Gemini 3.1 Pro | $2 | $12 | 2.4× pricier |
| Claude Sonnet 5 | $2 | $10 | 2× pricier |
| Gemini 3.8 Flash | $0.75 | $3.75 | 1.3× cheaper |
| Gemini 3.5 Flash-Lite | $0.30 | $2.50 | 2× cheaper |
| GPT-6 Luna | $0.10 | $0.50 | 10× cheaper |
Claude Haiku 4.5 supports a 200K token context window. That means in a single API call, you can send up to 200K tokens of input — including your prompt, system instructions, and any documents or code you want the model to process.
In practical terms, 200K tokens is approximately:
Compare all models' context windows on the context windows page.