How does token analysis consider reasoning tokens?

[aryoda]

I think some the models I have tested do use resoning but I cannot see the reasoning token consumption in the ocmonitor models breakdown:

I know that the token count for reasoning may be different for each model provider (some return the implicitly as "reasoning tokens", other just add them the output tokens).

Since reasoning tokens are charged like output tokens normally it would be helpful so see the reasoning token count too in the output.

Also: Additionally to the "Total Tokens" col I think it could make sense to also show the cached reads and write to make the "cost" col more understandable.

[Shlomob]

Hi
That is an interesting idea, i will investigate how reasoning tokens are saved by opencode and will add this feature in the next release.
Also, I added cache read/cache write to the model breakdown command (in the latest version , 1.04)

[aryoda]

Thanks a lot for your prompt implementation.

I have just tested it and think showing the cache reads/writes is the best way to handle this.

ocmonitor models

I propose to close this issue now after a cooling-off period...