Copilot Student Plan: 200 Credits Gone in a Single Day — This Wasn't the Case Before June 1

Hi everyone,

I'm a verified student on the GitHub Copilot Student plan and I want to share my experience because I think many others are facing the same issue.

Before June 1, 2026, I was regularly using Copilot Agent Mode in VS Code for my projects — sometimes doing large automation tasks, running agents, working with JSON files, and monitoring terminal output. I did this consistently for months without ever hitting any limit. The system worked great for my student workload.

Then on June 3, 2026 — just 2-3 days into the new billing cycle — I noticed that 180 out of my 200 monthly AI credits were already consumed. I hadn't done anything differently. Same workflow, same tasks, same Agent Mode I had always used.

The usage dashboard showed most of the consumption came from Claude Haiku 4.5 and GPT-5 mini — models I was already using before without any issue.

This is deeply frustrating because:

• I didn't change my usage habits at all
• The old system handled my workload perfectly fine for months
• The new credit system consumed nearly my entire monthly allowance in just 2-3 days
• As a student, I cannot afford to upgrade to a paid plan

I understand GitHub needs a sustainable business model, but the transition to token-based billing has made the Student plan practically unusable for anyone doing real development work with Agent Mode.

My questions for the GitHub team:

1. Will the Student plan credit limit be reviewed given this feedback from the community?
2. Is there any plan to offer more credits to verified students?
3. Why were students not given a proper warning about how drastically this would affect usage?

I'm not alone in this — many students are reporting the same issue. I hope GitHub takes this seriously and reconsiders the credit allocation for Student plan users.

Thank you.

I was using the same account in the whole month , but new udpation of git copilot is frustrated, My second day of devleopement limit have been reached ,
Actually new update of git copilot is a joke with Students

Yeah, I'd love having the deepseek v4 pro/flash available. They are pretty normal models and they are REALLY cheap, but 2$ allowance for GPT, Gemini or any of the big models out there? That's not gonna last a day

This may not be the place, i have an individual account, the same things are happening with us.
I noticed i cant select a model anymore.
The auto selection will select what ever it wants, and it will/did select a high cost model and that will/did eat up all the credits.
This can go from 0.33 to 6.x, whatever that means, all I know is what was lasting a month is now gone in under a week.
What is DeepSeek??
Is it a solution??
Yes I will be looking for that answer, money talks, so I will be talking to someone else!!

so this is an update of a real world use,
the question:
show code to convert a std::string to a std::filesystem::path
Auto routes based on your task and real-time system health and model performance. Learn More
Non-reasoning - Confidence 66%
Use this:
#include
#include
std::string s = "/tmp/example";
std::filesystem::path p = std::filesystem::path(s);
If you want the path from a string literal directly:
std::filesystem::path p = "/tmp/example";
And if you need to convert back to a string:
std::string str = p.string();
MAI-Code-1-Flash • 3.7 credits

same question to brave web browser
#include
#include

namespace fs = std::filesystem;

int main() {
std::string inputPath = "/home/user/file.txt";

// Method 1: Implicit conversion (copy initialization)
const fs::path path1 = inputPath;

// Method 2: Explicit construction
const fs::path path2(inputPath);

// Method 3: Assignment
fs::path path3;
path3 = inputPath;

return 0;

}
so what is the copilot really doing for me/us??

I got the same problem, I used the whole credit in 3 light tasks.

You should probably report this as a billing or usage-tracking bug too.

Your dashboard says you’ve used 200 / 200 AI credits, but when you look at the model breakdown, it only adds up to:

Auto: GPT-5.3-Codex: 34.47 credits
Auto: GPT-5 mini: 13.42 credits
Auto: Claude Haiku 4.5: 2.91 credits

That’s only 50.80 credits total, which means 149.20 credits are missing/unaccounted for.

i was used the same account in the whole month , but new udpation of git copilot is frustrated, My second day of devleopement limit have been reached ,
actually new update of git copilot is a joke with students

It's not a bug, it's a feature! Feature of making money😉

you nailed it

omg :c

truth😂

I dont understand my Student Plan anymore, it Shows Git Hub Pro but Copilot Free, so at the moment my Copilot Pro is also gone ? When I want to activate again it shows thats its not possible at the moment... So bad at this moment because I ride my thesis and need some change in my Plot codes wich is anoying to do myself... With a few changes my Free Plan is emty....

Yes, but they updated my crediets consumption shortly afterward.

mine is slowly increasing for some reason? seems like a delayed billing issue

For anyone reading, I just ended up switching to Kilo + Deepseek. It's so cheap. I put like $5 into it -- I've used up ~5 cents (US) in the same amount of work as my entire 200 github student credits.

Yeah, I totally recommend it.

I used to have the GitHub Student Pack, but suddenly I realized I had reached the Student Pack limit. So I put $10 into GitHub Pro to get more access to GitHub Copilot… but very soon, I hit the limit again! 😅

Then I tried DeepSeek and added another $5 to it. With Kilo Agent, I’ve been going deep into coding with it.

And honestly, this is what I’ve managed to do with just $8.44 so far:
981,398,058 tokens used
6,852 API requests
Total cost: just $8.44

Almost 1 BILLION tokens for $8.44?!
That’s honestly unbelievable! 🤯

Yah I think so too and where did you get my email form

Not sure about your question but I saw this thread on Google

Yes, I understand that the service has been heavily abused by bad actors. However, I believe GitHub/Microsoft should focus more on strengthening safeguards against abuse and exploitation rather than imposing such severe restrictions that legitimate users are left with barely any usable capacity. @github :)