Short answer
An API request costs (input tokens × input price + output tokens × output price) ÷ 1,000,000. With 2,000 input and 500 output tokens per request and 1,000 requests a day, Claude Sonnet 4.6 costs $0.00517 per request and $155 a month at APIVAI as of 2026-10-10, against $405 a month at Anthropic's official list price. The cheapest model in the table below costs $2.02 a month for the same workload (GPT-6 Luna). Change the numbers above to see your own workload.
What does this workload cost with each model?
2,000 input and 500 output tokens per request, 1,000 requests a day, 30 days. Sorted from cheapest to most expensive.
| Model | Per request | Per month | Official price per month | Saving |
|---|---|---|---|---|
| GPT-6 Luna | $0.0000672 | $2.02 | $13.50 | 85% |
| GPT-5.6 Luna | $0.000151 | $4.52 | $30.00 | 85% |
| Claude Haiku 5.5 | $0.00086 | $25.80 | $13.50 | — |
| GPT-6.1 Sol | $0.00136 | $40.80 | $270 | 85% |
| GPT-6 Sol | $0.00136 | $40.80 | $270 | 85% |
| GPT-5.6 Terra | $0.00151 | $45.30 | $300 | 85% |
| Claude Haiku 4.5 | $0.00172 | $51.60 | $135 | 62% |
| GPT-5.6 Sol | $0.00304 | $91.35 | $540 | 83% |
| Claude Sonnet 5.5 | $0.00345 | $103 | $270 | 62% |
| Claude Sonnet 5 | $0.00345 | $103 | $270 | 62% |
| GPT-5.5 | $0.00382 | $115 | $750 | 85% |
| Claude Sonnet 4.6 | $0.00517 | $155 | $405 | 62% |
| GPT-6 Astra | $0.00684 | $205 | $1,350 | 85% |
| Claude Opus 5.5 | $0.0069 | $207 | $540 | 62% |
| Claude Opus 5 | $0.00863 | $259 | $675 | 62% |
| Claude Opus 4.8 | $0.00863 | $259 | $675 | 62% |
| Claude Opus 4.7 | $0.00863 | $259 | $675 | 62% |
| Claude Opus 4.6 | $0.00863 | $259 | $675 | 62% |
| Claude Fable 5.1 | $0.0199 | $596 | $1,350 | 56% |
| Claude Fable 5 | $0.0199 | $596 | $1,350 | 56% |
How is the cost calculated?
Prices are per million tokens, separately for input (your prompt, the conversation so far, files and tool results) and output (the model's reply). The cost of one request is input tokens × input price ÷ 1,000,000 plus output tokens × output price ÷ 1,000,000. The monthly figure is the cost per request × requests per day × 30.
Three things change the real bill:
- Thinking tokens. Thinking is on by default and is billed as output, so a reply can use more output tokens than the text you see.
- Conversation history. Chat apps and coding agents resend the earlier messages with every request, so input grows as a conversation goes on.
- Prompt caching. A repeated prompt prefix can be read from cache at a much lower price. The calculator leaves caching out; see each model page for cache prices and a cached example.
How many tokens does a typical request use?
Rough sizes for planning. A token is about 4 characters of English text, or about 3/4 of a word; other languages and code often use more tokens per word.
| Request | Input tokens | Output tokens |
|---|---|---|
| A chat message with some history | 2,000 | 500 |
| One step of a coding agent (files and history resent) | 30,000 | 1,500 |
| Summarizing a 15-page document | 20,000 | 1,000 |
| Classifying a short text | 500 | 20 |
Your API responses report the exact counts (usage.input_tokens / usage.output_tokens in the Anthropic format, usage.prompt_tokens / usage.completion_tokens in the OpenAI format), and the Dashboard lists the tokens and cost of every request.
FAQ
How much does the Claude API cost per request?
It depends on the model and the size of the request. A 2,000-token prompt with a 500-token reply costs $0.00517 with Claude Sonnet 4.6 and $0.00172 with Claude Haiku 4.5 at APIVAI as of 2026-10-10.
Are input and output tokens priced the same?
No. Output tokens cost several times more than input tokens for every model, so long replies and thinking add up faster than long prompts.
Does the calculator include prompt caching?
No. It uses the normal input price. With caching, repeated input is billed at the much lower cache-read price, so real costs for agents and long conversations are usually lower than shown.
Is there a monthly fee or minimum spend?
No. APIVAI is pay-as-you-go from a prepaid balance: you top up from $10 and each request is deducted at the prices on the pricing page. There is no subscription.
Where do the official prices come from?
From the providers' published list prices per million tokens, which the price list is checked against every 6 hours.
Create an account, top up from $10 and check the real cost of each request in the Dashboard.