BETA
Tools · Budget Calculator
This is a free AI cost calculator for token-billed APIs. Set your request volume and token sizes to project what a month of inference costs, on any model in the index.
As of September 2026, the models in this calculator blend to a median of $0.80 per million tokens (our 70/30 weighting of published rates, not a price any provider quotes), with half between $0.25 and $2.22.
| Rank | Provider | Model | Per request | Per 1K | Per 1M |
|---|---|---|---|---|---|
| #1 |  | Gemma 2 2B Nano | $0.000050 | $0.05000 | $50.000 |
| #2 |  | Llama 3.1 8B Instruct text->text | $0.000065 | $0.06500 | $65.000 |
| #3 | inclusionai | Ling 3.0 Flash text->text | $0.000073 | $0.07350 | $73.500 |
| #4 |  | Qwen2.5-1.5B-Instruct Nano | $0.000075 | $0.07500 | $75.000 |
| #5 | | Llama 3 8B Instruct text->text | $0.000080 | $0.08000 | $80.000 |
| #6 |  | Granite 4.0 Micro text->text | $0.000090 | $0.09000 | $90.000 |
| #7 | | Llama 3.2 1B Nano | $0.000100 | $0.100 | $100 |
| #8 |  | Phi-3-Mini-4K (NIM) Nano | $0.000100 | $0.100 | $100 |
A free LLM cost calculator and AI spend estimator. Set a daily request count and the input and output tokens a typical request carries, and it projects per-request, monthly and at-scale cost. To count the tokens in a real prompt rather than estimating them, use the prompt calculator. Browse the AI model prices for the full rate sheet behind these projections.
A workload of 10 million tokens a month therefore costs about $8.00 on the median model and $0.20 on the cheapest paid one. 11 of the models priced here carry no per-token charge at all. The most expensive is OpenAI o1-pro at $285.00 per million tokens blended.
AI models bill per token, and a token is roughly four characters of English text. Every request is priced in two parts: the input tokens you send (system prompt, context, and message) and the output tokens the model generates. Providers publish these as per-million-token rates, and output is almost always the dearer side. A monthly bill is then just that per-request cost multiplied by your request volume. Estimating before you build matters because output length and request volume compound: a workload that looks cheap per request can dominate an infrastructure budget at production scale. If you do not know how many tokens your prompts actually carry, count a real one on the AI token calculator first and bring the numbers back here. How the blended rate is calculated →
A single request of 1,000 input tokens (roughly a page of context and a short question) and 300 output tokens (a few paragraphs of answer), priced on live rates at three points across the index:
| Model | Input, per 1M | Output, per 1M | This request |
|---|---|---|---|
| DeepSeek-R1-Distill-8BBudget tier · DeepSeek | $0.070 | $0.200 | $0.00013 |
| Mistral Large 3Median model · Mistral | $0.500 | $1.50 | $0.00095 |
| Kimi K3Premium tier · Moonshot AI | $3.00 | $15.00 | $0.00750 |
Multiply by your request volume to project a month: at 100,000 requests the same workload runs $13.00 on DeepSeek-R1-Distill-8B and $750 on Kimi K3. That gap, not the per-request figure, is what makes model choice a budget decision.
Pick a model, then enter your expected input and output token volumes per request and your daily request count. The calculator multiplies those token counts by the model’s per-million-token input and output rates to give a per-request cost, then scales that across your request volume to project monthly spend.
It is the product of three numbers: requests per month, tokens per request, and the model’s per-million-token rates. A workload of 1,000 requests a day at 2,000 input and 500 output tokens runs 30,000 requests and 75 million tokens a month, so the monthly bill moves by hundreds of dollars depending only on which model you pick. Set your own figures above to see the spread.
A blended price collapses a model’s separate input and output rates into a single figure using an assumed ratio of input to output tokens (commonly 70% input, 30% output). It is useful for quick apples-to-apples ranking, but for an accurate estimate you should price your actual input and output token volumes separately, which is what this calculator does.
No. Projections use each model’s standard flat per-token input and output rates. Prompt caching, batch endpoints and volume tiers can cut a real bill substantially, so treat the figure here as the undiscounted ceiling and check the provider’s billing page for the discounts your workload qualifies for.
Yes. It is free to use and needs no sign-up. Pricing comes from the same provider-published rates as the rest of Tokenando.