LLM Cost Calculator
What AI-assisted analysis costs per call and per month at your own token prices, how much of it is output, and how many calls match a subscription.
- Fresh input
- $0.02400
- Cached input
- $0.00000
- Output
- $0.01800
- Cost per call
- $0.0420
- Output share of cost
- 42.9%
- Per day
- $1.68
- Per month (30 days)
- $50.40
- Calls per month to match the subscription
- 476
Prices are yours to enter because model pricing changes often — check your provider's current page rather than trusting a number baked into a calculator. The output-share row is usually the surprise: output tokens are typically priced several times higher than input, so a verbose response costs more than a long prompt.
The arithmetic
cost per call = fresh input × input price
+ cached input × cached price
+ output tokens × output price
(prices quoted per million tokens)Simple enough that the interesting part is not the formula but which term dominates, and that is usually not the one people assume.
Output is where the money goes
Providers typically price output several times higher than input. Watch the output-share row: for a setup sending a large chart context and receiving a paragraph of analysis, output can still be the majority of the cost.
The practical lever follows directly. Shortening the response format — asking for a structured plan rather than an essay — often saves more than trimming the context does. If you are optimising the wrong end, the calculator will show it.
Caching changes the shape
A trading assistant tends to send a large, stable block on every call: a system prompt, instrument definitions, your rules. Where the provider offers a reduced rate for repeated input, moving that share into the cached field can change the monthly figure substantially.
Set the cached share to 0% and then to something realistic for your setup. The gap is the value of structuring your context so the stable part is actually stable, rather than regenerated slightly differently each call.
Your key or hosted credits
The break-even row answers the narrow version of the question: how many calls a month before a subscription is cheaper than paying per token.
It does not answer the whole question. Using your own key means the provider relationship, rate limits and spend are yours, and the key stays on your machine alongside your exchange credentials. A subscription trades that for not thinking about it. CoTrading is free with your own model key and sells hosted credits only as the convenience option — the arithmetic above is the honest input to which you want.
What this leaves out
- Retries and failed calls. They cost the same as successful ones.
- Context growth within a conversation. A long session resends accumulated history each turn, so the input figure is not constant.
- Tool-call round trips. A single user question that triggers several tool calls is several billable calls.
That last one matters for an MCP setup in particular: a model answering "how is my portfolio doing" may fetch balances, positions and several tickers, each round trip adding tokens. See rate limits in an MCP trading setup for the same pattern viewed from the exchange side.
FAQ
Why do I have to enter the token prices myself?
Because model pricing changes often enough that a table baked into a page would be wrong within weeks, and a wrong price here produces a confidently wrong monthly figure. Read the current numbers from your provider’s pricing page and enter them; the result is then yours rather than approximately somebody else’s from a while ago.
Why is output so much more expensive than input?
Providers generally price output tokens several times higher than input, because generating them is more computationally expensive than reading them. The practical consequence for a trading assistant is that a verbose response costs more than a long chart context — so trimming the answer format is often a bigger saving than trimming the prompt.
What is cached input and when does it help?
Many providers offer a reduced rate for input that repeats across calls, such as a fixed system prompt or a document you keep referring to. It helps when a large, stable block of context is sent on every call — which describes most assistant setups. It does not help when each call sends genuinely new material.
Should I use my own key or a subscription?
It comes down to volume, which is what the break-even row is for: it shows how many calls per month a subscription has to displace before it is the cheaper option. Below that, your own key wins; above it, the subscription does. Everything else — convenience, rate limits, wanting keys to stay on your machine — is a separate question the arithmetic does not settle.