See what a chat costs
English
Summary
This page shows how to see an estimate of what a chat has spent, under the message box. It helps when you pay a provider by use and want to know which chats are expensive. You need Delorun running with a model provider (Model providers). The glossary describes the cost meter. The meter is off until you turn it on, and it makes no request of its own: it adds up the usage the provider already reported.
Contents
Turn it on
Open Plugins in the sidebar and open Delorun studio. Under Components, find Cost meter and turn it on; it starts at once, with no restart. You should see a figure under the message box in every chat, after the token count.
The first-run guide's Recommended plugins step offers the same switch.
What it shows
The box shows what this chat cost, such as $0.42. When a chat has started subagents or teammates, it shows two figures, $0.42 · $0.61 with agents: this chat alone, then this chat with the chats it started.
Select it to see the breakdown: one row for each model, with its requests, tokens and cost. Each row opens into input, cache write, cache read and output tokens. With agents, a row for this chat comes first, then one for each agent it started, costliest first.
The figure is an estimate. A request with no price in the table is listed as such rather than guessed.
Prices
Prices come with Delorun and are checked against each provider's published prices when a release is prepared. Long prompts and DeepSeek's peak hours are priced as the providers do. Nothing leaves your computer to work them out.
Edge cases
- If a chat runs a model the price table does not know, then the breakdown lists its requests as unpriced and says so.
- If the figure differs from your invoice, then check What it does not count. The invoice is the authority.
- If a price changed after your release, then override it as described under Go further.
- If you reload, resume or fork a chat, then the total is the same, because it is worked out from the chat's saved history.
- If you turn the meter off, then the figure disappears. Your chats keep their history.
- If the settings you give the plugin are malformed, then the plugin stops at load. Fix them or remove them.
Go further
An entry under prices in the plugin's settings replaces the price table's entry for that model, or adds a model. Each entry names a model (a prefix of the model id; the longest matching prefix wins) and four rates in US dollars per million tokens: input, output, cacheRead and cacheWrite. Add provider to limit the entry to one provider, longContext for a higher rate above a prompt size, or peak for rates in the provider's busy hours. Changing your harness explains where plugin settings live and how to undo a change.
What it does not count
Batch prices, fast mode, one-hour cache writes, discounts and Chinese holidays are not modelled.
Next
- Start chatting: Daily use.
- Choose cheaper or other models: Model providers.
Your Delorun home
Your Delorun home: where delorun keeps your data, how to move it with DELORUN_HOME, profiles, why ~/.dsh stays untouched, and how to fix a home or secret file that other users can read.
Read the model's thinking word by word
Read the model's thinking word by word: turning on the Thinking tab, the band and the page, choosing what to read, the pace and the delay, and where your choices are kept.