HuaRenCa
Back to Forum
Community

How do I fix Gemini Pro's Argon frequently hitting the limit and pausing?

tian田峰
tian田峰

2 weeks ago

Once it hits the limit and pauses, I can only use Flash Lite, and the quality is terrible—uploaded text and images can't be recognized correctly, and the hallucinations are obvious. When I point it out, it apologizes repeatedly, but keeps making the same mistakes.

Does Argon use a lot of tokens? Or are there any tricks for using it? I have several conversations on different topics, and I ask follow-up questions under each one, which makes them easy to find. But does this make the context too long?

I'd appreciate advice from the experts. Thanks~

7
638

Comments (7)

Your avatar
Sign in to comment
dexyfun
dexyfun2 weeks ago

You didn't specify whether it's the free quota or the paid quota that's been exceeded—these two situations call for completely different approaches.

笑儒
笑儒2 weeks ago

You've piled follow-up questions on several topics into a single thread—how exactly do you interleave them? Do you follow up on each one right after the preceding context, or do you only switch over at certain points? This directly determines whether the context accumulates linearly or gets stretched out over and over, and it also affects whether you conclude the issue is the number of topics or the depth of a single topic.

GoldenDreams
GoldenDreams2 weeks ago

Asking follow-up questions on multiple topics within a single conversation is certainly convenient for retrieval, but each follow-up can drag unrelated earlier content into the context, so it's easy to burn through your quota the more you use it.

方舟集运
方舟集运1 week ago

Is it possible that the problem isn't Argon itself, but rather that the uploaded images or long text are using up a large amount of the quota first?

狼人善牧
狼人善牧1 week ago

I tend to think the core issue here is context management rather than the model itself: when you pile multiple topics into a single conversation, every follow-up question has to carry the entire history again, so token usage balloons quickly with each turn—hitting the limit is just the outcome. If you really want to control the risk, you have to accept a drop in retrieval convenience and either split things up by topic or start a fresh conversation periodically.

刘德华888
刘德华8881 week ago

Start a new conversation with just a single topic and compare usage under the same prompt to see whether context accumulation is the cause. If limits are still hit frequently after splitting, then consider adjusting how you use it.

Terry
Terry1 week ago

Keep asking follow-up questions in the same conversation and the history keeps growing, so your quota gets used up faster and faster.