What will your LLM calls actually cost?
Pick your models, enter a realistic request size, and see API spend per request and per month side by side — using 30 models' published list prices, updated Oct 15, 2026.
Token counts are approximated at ~4 characters per token. Use your provider's tokenizer for exact billing.
| Model | Cost / request | Cost / month | Tokens / month |
|---|
Monthly spend by model
How it works
- Describe one request. Input tokens are your prompt, system message, retrieved context and conversation history. Output tokens are what the model writes back.
- Set your volume. Requests per day is multiplied by 30 to get a monthly figure.
- Compare. Costs are list price × tokens. The cheapest option updates as you type.
Popular comparisons
Featured models
| Model | Vendor | Input / 1M | Output / 1M | Context | Max output |
|---|---|---|---|---|---|
| GPT-6.1 Sol | OpenAI | $2 | $10 | 1.1M | 128K |
| Claude Sonnet 5.5 | Anthropic | $2 | $10 | 1M | 128K |
| Gemini 3.8 Flash | $0.750 | $3.75 | 1.0M | 65.5K | |
| DeepSeek-V4.1-Flash | DeepSeek | $0.300 | $1.20 | 1M | 384K |
Who this is for
Builders choosing a default model
You have a working prototype and need to know what it costs at 10× the traffic before you pick a provider to build around.
Teams under a budget cap
You need a defensible number for a planning doc: per request, per month, per thousand users.
Tinkerers comparing releases
A new model ships, the price looks low, and you want the real per-request number at your prompt size — not the headline per-million rate.
Frequently asked questions
Is SelfBench free?
Yes. Every calculator and the full model catalog are free and need no account. Pro ($19/month or $149/year) adds a pricing API, bulk estimates and dataset exports; the free tools stay free.
Where do the prices come from?
From each vendor's own published pricing page. Every model page links to the exact source and shows the date we last verified it. We do not scrape resellers or estimate prices from third parties.
How accurate is the cost estimate?
The arithmetic is exact for the numbers you enter: tokens × list price. The uncertainty is on the input side — your real token counts, caching hits, batch discounts and negotiation. Treat the output as a solid planning number, not an invoice.
Do you store my prompts?
No. The calculators run entirely in your browser. Text you paste into the prompt box never leaves your device. See the privacy policy for the short list of things we do store (and why).
Which models are covered?
Currently 30 models from 9 vendors: Alibaba Cloud (Qwen), Anthropic, DeepSeek, Google, Meta, Moonshot AI, OpenAI, Zhipu AI, xAI. We add new releases as vendors publish prices, and every change is listed in the data changelog.
Can I use the data in my own tool?
Yes — that is what Pro is for. The pricing API returns JSON, and you can export the full dataset as CSV. Free browsing on this site always stays free.
Do you handle tax and invoices?
Our payment provider, Waffo Pancake, acts as the Merchant of Record: it processes the payment, issues the invoice and handles applicable sales tax or VAT.
How do I cancel?
Ask us and we will cancel it, or use the customer portal link we send after purchase. Cancelling stops future renewals; access continues to the end of the paid period. Refunds are covered by the refund policy.
Understand the numbers
How to estimate LLM API cost (and where the estimate goes wrong)
The arithmetic is the easy part. Choosing the inputs is where estimates fail.
Cached input pricing across LLM APIs
The single biggest lever on input-heavy bills — and the one most estimates ignore.
LLM pricing traps: tiers, promos, off-peak rates and cache writes
The headline rate is a starting point, not the bill.
Context window planning: what a 1M-token window actually costs
A big context window is a capacity limit, not a budget.
Need this inside your own pipeline?
Pro adds a JSON pricing API, CSV export and bulk cost estimation. $19/month, cancel any time.
See Pro pricing Read the API docs