Model Dashboard
The Model Dashboard shows how you use models in model chats: how many questions you asked, how many tokens they used, what that is estimated to cost, and how each model compares on speed and price. Agent and swarm runs aren't included.
To open the screen, go to Agent Ops → Model Dashboard.
Before you start
The dashboard shows your own model chats only. If you haven't asked a question in a model chat yet, you see No model chat usage yet.
Choose a time range
Select 7 days, 30 days or 90 days at the top of the screen. The dashboard opens on 30 days.
The range changes the headline, the charts and the Model detail table. The headline tells you how much you spent in the range, how many questions you asked and across how many models. Updated shows when the figures were read, in your own time zone.
Summary
| Card | What it shows |
|---|---|
| Today | The estimated spend so far today, by your own calendar day. |
| This month | The estimated spend this month, and how it compares with last month. |
| Forecast | A straight-line forecast of this month's spend, based on the days so far. |
| Per question | The average tokens per question, with the average response time underneath. |
Today, This month and Forecast are calendar figures. They don't change when you choose a different time range.
Charts
- Daily spend shows the estimated spend for each day in the range, and marks the day with the highest spend. Point at a bar to see the day and the amount.
- Where the money goes ranks the models by spend, with how many times each was used. Models beyond the top five are grouped together.
- Speed vs cost compares the average response time of your most-used models. When one model answers faster and for less than another you use often, the chart suggests sending more questions to it.
Model detail
The Model detail table lists each model you used.
| Column | What it shows |
|---|---|
| Model | The model name. |
| Uses | How many questions went to the model. |
| Tokens | Tokens used. |
| In / 1M | The list price per million input tokens. estimated means the model has no published price, so a blended rate is used. |
| Out / 1M | The list price per million output tokens. |
| Cost | The estimated cost. |
| Latency | The average response time. |
| Last used | When the model was last used. |
The table shows the top five models first. Select Show all, followed by the number of models, to see every model, and Show fewer to collapse the list again.
How costs are worked out
Costs are worked out from the recorded token counts at published list prices. They are estimates, not an invoice, and your provider may bill differently. The note under the table explains any models that aren't listed, and says when only the most recent chats could be read, so the totals are a minimum.
Troubleshooting
| Message | What to do |
|---|---|
| Usage could not be loaded | Reload the page to try again. |
| Model Dashboard is unavailable | Model chat usage isn't recorded for your account. Contact support. |
| Model Dashboard is temporarily unavailable. | Reload the page after a moment. If the problem continues, contact support. |