Burst should refill with idle time
If I don’t use chat for a while, I’d expect my “burst” to refill, at least a bit. A once-per-month refill of burst really doesn’t fit my usage pattern, but if I’m low (or empty) on burst, I’d love to just hold off for a while in order to give myself more usage later in the day. More transparency on what the balances actually represent would be helpful too - so much vague “dollar signs” and “percent remaining”, not really enough information to let me know what I can and can’t do
Log in to comment and vote
Comments5
Mark Florkowski
The “base” bucket refills on a 4-hour cadence. The burst overage is a larger bucket that refills monthly when your subscription renews.
We are still tuning the bucket sizes for how they feel best. It looks like you pretty much exclusively use Gemini 3.1 pro which is among the most expensive models, and thus far have already used more than the $50 you have paid. We do allow this, but there is a limit to how much we are able to subsidize, and to be frank it is nowhere near that of what multi-billion dollar companies like Google can do.
Purple Kangaroo
Well crap, if I'm using more than what I'm paying for, you definitely need to tune it, yeah? Go ahead and cap me, I'm using this thing for the good ui and model selection, not to take advantage.
My ideal would be something that lets me have complete control of my “budget” so that I can stop myself from burning through it all at a pace I can't sustain. That's my goal here - a cap that I can control to limit my spend, not saving on cost per call.
Considering switching to byok anyway (for a single bill that covers apis as well) which might moot this, but still the opaqueness of the usage meter has been something I've thought about fairly constantly as I've tried t3chat
Mark Florkowski
Appreciate it, and re: capping limits, the way that the new system works contains usage enough to not put us underwater, we’re trying to find ways that can make it go further for our users though
| complete control of my “budget” so that I can stop myself from burning through it all at a pace I can't sustain.
Would some form of indication in the chat that “you are getting close to x% of your 4-hour quota” or “hey that last message was kind of expensive” be helpful for this goal?
| Considering switching to byok anyway
Honestly I would recommend it if you understand how API keys work, which it clearly sounds like you do. At the individual scale costs are pretty manageable and you can also benefit from some individual grants that many of the major labs offer, while still using our UI.
Purple Kangaroo
I thought it already did that? Like, I’m getting “consider starting a new thread” messages already, but I seem to have a different opinion on what that threshold should be.
Right now you output how many tokens were in the response. Would it be possible to output a split of: Input tokens, Reasoning tokens, Response Message Tokens? Seeing the input tokens grow over time could drive home the “this is getting kind of expensive”, while seeing the reasoning tokens being the majority of the response could serve as a reminder to turn off high thinking
“that was kind of expensive”, in general, does sound like a good feature, it’s just a matter of conveying that expense so that I don’t dismiss the popup without consideration, because ultimately I know I’m just trying to work around my own bad habits with statistics (a need that won’t go away with byok)
Ivory Control
I think that you should just move the base and burst usage bars to the main chat window. From a user’s perspective, this is the superior solution. In fact, I even made a request for it.