AI API cost calculator
What will it cost each month? Enter your volume and see 14 models side by side, from list prices.
10,000 requests a month, 1,000 tokens in and 250 out each: 10 of 14 models have a supplied price.
| Model | Input, $ per 1M tokens | Output, $ per 1M tokens | Per request | Per month |
|---|---|---|---|---|
| GPT-6 LunaOpenAI | $0.10 | $0.50 | $0.0002 | $2.25 |
| DeepSeek V4.1 FlashDeepSeek | $0.30 | $1.20 | $0.0006 | $6.00 |
| Gemini 3.8 FlashGoogle | $0.75 | $3.75 | $0.0017 | $16.88 |
| GLM-5.3Zhipu (Z.ai) | $1.40 | $4.40 | $0.0025 | $25.00 |
| Grok 4.7xAI | $2 | $6 | $0.0035 | $35.00 |
| Claude Sonnet 5.5Anthropic | $2 | $10 | $0.0045 | $45.00 |
| GPT-6.1 SolOpenAI | $2 | $10 | $0.0045 | $45.00 |
| Claude Opus 5.5Anthropic | $4 | $20 | $0.0090 | $90.00 |
| Claude Fable 5.1Anthropic | $10 | $50 | $0.0225 | $225.00 |
| GPT-6 AstraOpenAI | $10 | $50 | $0.0225 | $225.00 |
| Kimi K3Moonshot AI | Not supplied | Not supplied | Unknown | Unknown |
| Mistral Medium 3.5Mistral | Not supplied | Not supplied | Unknown | Unknown |
| Muse Spark 1.3Meta | Not supplied | Not supplied | Unknown | Unknown |
| Qwen3.8 MaxAlibaba/Qwen | Not supplied | Not supplied | Unknown | Unknown |
How the estimate is made
Assumptions
monthly cost = requests × (input tokens × input price + output tokens × output price) ÷ 1,000,000
It leaves out prompt caching, reasoning tokens, retries, batch discounts, tiered or long-context pricing, tool and search fees, image, audio and video units, and taxes. Real bills are usually higher than this estimate when a model reasons before it answers. Treat the result as a way to compare models, not a quote.
Frequently asked questions
How is the monthly cost calculated?
Requests a month multiplied by the sum of your input tokens times the input price and your output tokens times the output price, divided by one million. Prices are USD list prices per million tokens.
Why is my real bill higher than the estimate?
The estimate leaves out reasoning tokens, retries, prompt caching, batch discounts, tiered or long-context pricing, tool and search fees, image, audio and video units, and taxes. Models that reason before answering often use far more output tokens than the reply you see.
Does the cheapest model do the job best?
Not necessarily. The table is ordered by cost only. We publish no quality benchmark for these models, so a low price says nothing about whether the model can do your task. Test the cheapest candidates on your own prompts.
Why does a model show an unknown cost?
No list price was supplied for it. We show unknown rather than estimate one, and the model stays at the bottom of the table until a price is supplied.
Are the prices verified?
Not yet. The editorial team supplied them on 2026-10-14 and we have not re-checked each one against the vendor page. Check the provider's pricing page before you budget.