AI pricing
What you actually pay when AI generates indicators for you, and why it's the provider's rate, not ours.
StrategyTune uses third-party AI models (Anthropic, OpenAI, Google) to generate custom indicator code. Every dollar you spend on AI inside StrategyTune is passed straight through to the model provider — we don't add a markup. The rates you see in the app's AI Models tab are the same rates published by the provider.
How AI is paid for. AI usage is billed against your monthly subscription credits first, then against any purchased top-up credits. If both are exhausted, AI features stop until your subscription renews or you buy more credits. There is no overage charge on top of what you already paid for.
What you pay for
Models bill on three things: input tokens (your prompt), output tokens (the response), and cache. Output is the most expensive — typically several times the input rate.
Tokens
A token is roughly ¾ of an English word, give or take. A short prompt is a few hundred tokens; a long indicator with examples can be a few thousand. Rates are quoted per 1 million tokens (per M).
Caching
Caching applies only to input tokens. When the start of a prompt repeats across turns — typically the system instructions and tool definitions — the provider can store it the first time and reuse it on later turns at a discount. Output tokens are never cached.
- Cache read — discounted rate when a previously stored prompt prefix is reused. Typically about 10× cheaper than regular input.
- Cache write — rate when a prompt prefix is stored for the first time. Anthropic charges a small premium (~25% above input). OpenAI and Google have no extra fee here — the first turn is billed at the regular input rate.
Where to see live rates
Rates can change when providers update them. The current numbers always live in the app:
- Open StrategyTune.
- Click your avatar → Billing & AI.
- Switch to the AI Models tab.
That table shows input, output, cache-read, and cache-write rates for every model we currently expose, sourced from the provider directly.
How to keep costs down
- Pick the smallest model that gets the job done. Cheaper models are great for tweaks to existing indicators; reserve the flagships for greenfield generation or hard logic.
- Reuse a chat thread when iterating on the same indicator. Caching kicks in on consecutive turns and drops input cost dramatically.
- Trim the conversation. Long back-and-forth means more input tokens replayed every turn. Start a fresh chat for a new indicator instead of piling onto an old one.
Something missing or wrong? Email support@strategytune.com.