| Vendor | Model | Input / 1M | Output / 1M | Cache hit / 1M | Monthly cost ▲ | Per 1,000 requests |
|---|
Figures in the table are converted into the currency you selected (original list prices are in the source table below). Click a column header to sort. The three cheapest models are highlighted in green.
Official prices and sources
| Vendor | Model / API ID | List price (input / output / cache, per million tokens) | Notes | Source |
|---|
How to use this calculator
- Estimate requests per day, plus average input and output tokens per request (the
usagefield in API responses has exact counts). - If your prompts share a stable prefix (system prompt or a long document), turn on prompt caching and enter a hit rate — costs usually drop a lot.
- Choose a display currency and, if you need yuan, the exchange rate. To see costs after scheduled increases, set “Pricing as of” to December 1, 2026 or January 1, 2027.
- Click the “Monthly cost” header to sort and pick a model that fits your budget.
How the token cost is calculated
- Monthly cost = requests per day × days per month × (input tokens × (1 − cache-hit rate) × input price + input tokens × cache-hit rate × cache-hit price + output tokens × output price) ÷ 1,000,000.
- Prices are each vendor's official standard (non-batch) list rates. OpenAI uses the short-context rate (input ≤272K). Anthropic Haiku 5.5 uses the prompt ≤100K rate. Alibaba Cloud uses the Beijing-region list price. DeepSeek uses the peak-hour rate.
- If a vendor does not publish a cache-hit price, the calculator uses the regular input price. Cache-write fees, web-search or tool surcharges, long-context uplifts, batch discounts, and limited-time coupons are not included.
- Price data lives in prices.js so it can be updated independently of the page.
FAQ
What is Gemini 4 Carbon pricing, and when does the promo end?
On Google's official pricing page, paid-tier Gemini 4 Carbon is $2 per million input tokens and $10 per million output tokens through November 30, 2026. On December 1, 2026 the price becomes $4 input and $20 output.
What is the Claude Fable 5.2 API price?
On Anthropic's official pricing page, Claude Fable 5.2 is $8 per million input tokens, $40 per million output tokens, and $0.20 per million for cache hits.
How do I use this AI API cost calculator?
Enter requests per day, input and output tokens per request, and an optional cache-hit rate. The table estimates monthly cost and cost per 1,000 requests for each model, converted into USD or CNY. Switch “Pricing as of” to see costs after scheduled price changes.
How does this LLM API pricing comparison 2026 work?
Prices are taken from each vendor's official pricing page and listed per million tokens for input, output, and cache hits. OpenAI figures use the short-context rate (input ≤272K). Anthropic Haiku 5.5 uses the prompt ≤100K rate. Alibaba Cloud uses the Beijing region list price. DeepSeek uses the peak-hour rate. Batch discounts, cache-write fees, tool and search surcharges, and long-context uplifts are not included.
How is monthly token cost calculated?
Monthly cost = requests per day × days per month × (input tokens × (1 − cache-hit rate) × input price + input tokens × cache-hit rate × cache-hit price + output tokens × output price) ÷ 1,000,000.
Are thinking or reasoning tokens billed on the API?
Yes. Vendors such as Google state that the output price includes thinking tokens. When you turn on extended reasoning, actual output tokens are often much higher than the visible answer length, so raise the output-token estimate accordingly.
Disclaimer
This site is not affiliated with the vendors listed above. Prices are taken from each vendor's official page (see the date at the top of this page). Vendors can change prices at any time; your invoice and the official pricing page are authoritative. The exchange rate is a sample value. Results are estimates only. All calculations run locally in your browser; nothing you type is sent or stored.