priced twice.
AI Token & LLM API Cost Calculator
Enter your prompt size, answer size and daily request volume to see the monthly input cost, monthly output cost and total monthly cost for eight 2026 models — plus the cheapest alternative you could switch to.
Private · Local · No account
- Input, output and total cost
- Cheapest-model saving badge
- Private, in your browser
Prompt in, answer out, bill at the end of the month.
Input tokens and output tokens are billed at different rates — output usually costs two to five times more. This calculator splits both, prices them per million tokens across eight models, and flags the switch that saves the most.
Open the calculator- Input, output and total cost
- Cheapest-model saving badge
- Private, in your browser
AI token & LLM API cost
Privacy First: Your usage data never leaves your browser.GPT-5.5 costs $240.00 a month here; Qwen3.7 Flash costs $1.98.
| Model | Provider | Input $ / 1M tokens | Output $ / 1M tokens | Monthly input cost | Monthly output cost | Total monthly cost |
|---|---|---|---|---|---|---|
| Qwen3.7 FlashCheapest | Alibaba | $0.10 | $0.13 | $1.20 | $0.78 | $1.98 / month |
| DeepSeek V4 Flash | DeepSeek | $0.14 | $0.28 | $1.68 | $1.68 | $3.36 / month |
| Gemini 2.5 Flash | $0.30 | $2.50 | $3.60 | $15.00 | $18.60 / month | |
| Gemini 3.8 Flash | $0.75 | $3.75 | $9.00 | $22.50 | $31.50 / month | |
| Claude Opus 5 | Anthropic | $5.00 | $25.00 | $60.00 | $150.00 | $210.00 / month |
| GPT-5.5Your pick | OpenAI | $5.00 | $30.00 | $60.00 | $180.00 | $240.00 / month |
| Claude Fable 5.1 | Anthropic | $10.00 | $50.00 | $120.00 | $300.00 | $420.00 / month |
| GPT-6 Astra | OpenAI | $10.00 | $50.00 | $120.00 | $300.00 | $420.00 / month |
Sorted cheapest first. Every row bills the same 15,000 requests a month — only the per-token price changes, so the ranking is identical for any volume.
- 800 tokens prompt + 400 tokens output per request → 800 input and 400 output tokens.
- 500 requests a day × 30 days = 15,000 requests and 18,000,000 tokens a month.
- Monthly input cost = 12,000,000 prompt tokens ÷ 1,000,000 × $5 = $60.00.
- Monthly output cost = 6,000,000 output tokens ÷ 1,000,000 × $30 = $180.00.
- Output tokens usually cost 2× to 4× more than input tokens, so long answers move the total more than long prompts.
- Prices are the published per-1-million-token rates; batch, cache and volume discounts are not applied.
- All calculations run locally in your browser.
Privacy First: Your usage data never leaves your browser. Estimates use list prices only — cached inputs, batch mode and long-context premiums can change your real bill.
How to Calculate LLM API Costs
Every LLM API bills two separate quantities: input tokens (the prompt you send) and output tokens (the completion the model generates). Each has its own price per 1 million tokens, and output is always the expensive side — you are paying for text that is generated one token at a time instead of read in a single pass.
That ratio is the whole game. On GPT-6 Astra and Claude Fable 5.1 the output rate is 5× the input rate ($50.00 against $10.00 per million). On DeepSeek V4 Flash it is only 2× ($0.28 against $0.14). A workload with long prompts and short answers therefore favours models with a low input price, while a chatty assistant with short prompts and long answers is dominated by the output price alone.
Convert words to tokens at 1 word = 1.33 tokensif your provider only reports word counts, multiply a day's volume by 30 days for a monthly figure, and keep the full precision until you display the result. Prices here are list rates per 1 million tokens: cached-input discounts, batch discounts and long-context premiums are not applied, so treat the number as a planning ceiling rather than an invoice.
The result is an estimate for budgeting, not financial advice or a provider quote.
The formulas
Monthly input tokens = prompt tokens per request × daily requests × 30
Monthly input cost = (Monthly input tokens ÷ 1,000,000) × input price per 1M tokens
Monthly output cost = (Monthly output tokens ÷ 1,000,000) × output price per 1M tokens
Total monthly cost = monthly input cost + monthly output cost
Tokens from words = words × 1.33
Worked example: 800 prompt tokens and 400 output tokens at 500 requests a day for 30 days is 12,000,000 input tokens and 6,000,000 output tokens a month. GPT-5.5 at $5.00 input and $30.00 output per million costs $60.00 + $180.00 = $240.00 a month. Qwen3.7 Flash at $0.10 and $0.13 costs $1.20 + $0.78 = $1.98 a month for the identical workload.
Cheapest LLM APIs in 2026
Budget models win on two levers: a low input rate for long prompts and a low output rate for long answers. DeepSeek V4 Flash and Qwen3.7 Flash undercut the frontier tier by roughly 50 to 100 times — $0.10 to $0.14 per million input tokens against $5.00 to $10.00 — while still serving production traffic at millions of tokens a day. For bulk classification, extraction, summarisation and high-volume chat, that gap compounds fastest because the workload is pure volume.
Google's Gemini 2.5 Flash ($0.30 input) and Gemini 3.8 Flash ($0.75 input) sit in the middle: still an order of magnitude below the frontier tier, with a wider quality margin for reasoning-heavy tasks. The premium models — GPT-6 Astra and Claude Fable 5.1 at $10.00 input and $50.00 output, plus GPT-5.5 and Claude Opus 5 at $5.00 input — earn their price on hard reasoning, long context and agent work, not on raw token throughput.
| Model | Provider | Input $ / 1M tokens | Output $ / 1M tokens | Output premium |
|---|---|---|---|---|
| Qwen3.7 Flash | Alibaba | $0.10 | $0.13 | 1.3× |
| DeepSeek V4 Flash | DeepSeek | $0.14 | $0.28 | 2.0× |
| Gemini 2.5 Flash | $0.30 | $2.50 | 8.3× | |
| Gemini 3.8 Flash | $0.75 | $3.75 | 5.0× | |
| Claude Opus 5 | Anthropic | $5.00 | $25.00 | 5.0× |
| GPT-5.5 | OpenAI | $5.00 | $30.00 | 6.0× |
| GPT-6 Astra | OpenAI | $10.00 | $50.00 | 5.0× |
| Claude Fable 5.1 | Anthropic | $10.00 | $50.00 | 5.0× |
List prices per 1 million tokens as registered for this calculator; the output premium is the output rate divided by the input rate. The live comparison table above re-prices the same eight models against your own prompt, answer and request volume. Confirm current rates on the provider's pricing page before you budget.
Frequently asked questions
How do you calculate LLM API cost per month?
Multiply your average prompt tokens per request by your daily requests and by the days in a month to get monthly input tokens, then divide by 1,000,000 and multiply by the model's input price per million tokens. Do the same with output tokens at the output price. The monthly input cost plus the monthly output cost is your total monthly bill. Example: 800 prompt tokens and 400 output tokens at 500 requests a day is 12,000,000 input tokens and 6,000,000 output tokens a month — on GPT-5.5 at $5.00 input and $30.00 output per million that is $60.00 plus $180.00, or $240.00 a month.
Are output tokens more expensive than input tokens?
Yes. Every model in this calculator charges more for output tokens than for input tokens — from about 2x on DeepSeek V4 Flash and Qwen3.7 Flash up to 5x on GPT-6 Astra and Claude Fable 5.1. Input tokens are prompt text the API mostly reads; output tokens are text the model generates token by token, which is the slower, more expensive work. That is why long, verbose answers move your bill far more than long prompts do.
Which LLM API is cheapest in 2026?
Qwen3.7 Flash is the cheapest model in the 2026 registry at $0.10 per million input tokens and $0.13 per million output tokens, followed closely by DeepSeek V4 Flash at $0.14 input and $0.28 output. For the default workload of 800 prompt tokens, 400 output tokens and 500 requests a day, Qwen3.7 Flash costs about $1.98 a month against $240.00 on GPT-5.5 — a saving of roughly $238 a month. Google's Gemini 2.5 Flash and Gemini 3.8 Flash sit in the middle at $0.30-$0.75 input, while GPT-6 Astra, Claude Fable 5.1, GPT-5.5 and Claude Opus 5 are the premium frontier tier.