Version 0.0.344 · Commit c87f222 - Possible bug in premium request counting (FIXED IN 0.0.345)

Describe the bug

After upgrading to Version 0.0.344 (Commit c87f222), I noticed that executing a single task/prompt consumed roughly 20% of my premium requests (about 60 requests). This behavior did not occur in previous versions.

Affected version

Version 0.0.344 · Commit c87f222

Steps to reproduce the behavior

Upgrade to Version 0.0.344.
Execute a prompt that runs a long task.
Observe the remaining premium request count during the same execution, it decreases multiple times.

Expected behavior

Only one premium request should be counted per prompt execution, regardless of task duration or complexity.

Additional context

Total usage est: 69 Premium requests
Total duration (API): 10m 12.5s
Total duration (wall): 1h 9m 4.1s
Total code changes: 1459 lines added, 188 lines removed
Usage by model:
unknown 0 input, 7.4m output, 0 cache read, 0 cache write
(Est. 69 Premium requests)

Same issue here:
one task but 41 requests.

Total usage est: 41 Premium requests
Total duration (API): 4m 53.9s
Total duration (wall): 28m 26.3s
Total code changes: 44 lines added, 11 lines removed
Usage by model:
unknown 0 input, 1.8m output, 0 cache read, 0 cache write
(Est. 41 Premium requests)

claude-sonnet-4.5 (1x)

same here, both Haiku and Sonnet.

0.0.344
Commit: c87f222

Same here for claude-sonnet-4.5 (1x):

● Total usage est: 29 Premium requests
Total duration (API): 10m 51.0s
Total duration (wall): 33m 0.1s
Total code changes: 1528 lines added, 30 lines removed
Usage by model:
unknown 0 input, 3.9m output, 0 cache read, 0 cache write (Est. 29 Premium requests)

Same here, one prompt that usually took 2-3 premium requests took 29 this time. Also input, cache read/write all says 0. Something seems broken.

same here.

Hey folks -- attempting to reproduce on my end so we can investigate this.

For people who are affected by this: do you see that many premium requests being burned on the github.com Copilot usage page? I suspect this is just a client-side rendering bug and not actually spending from your premium request usage

Copilot Usage shows that it is consuming my premium requests. A single prompt of haiku-4.5 in the CLI spent ~10.89 (which is 33 x 0.33) premium requests.

2025-10-18 copilot copilot_premium_request 29 requests

My daily report does show 29 counted requests.

Hey folks -- attempting to reproduce on my end so we can investigate this.

For people who are affected by this: do you see that many premium requests being burned on the github.com Copilot usage page? I suspect this is just a client-side rendering bug and not actually spending from your premium request usage

Yes, it also shows that many(41) premium requests on the github.com Copilot usage page

same here

0.0.344
Commit: c87f222

A shorter session, the "usage by model" stats seem wrong? Using Claude Sonnet 4.5:

 Total usage est:       33 Premium requests
   Total duration (API):  3m 42.2s
   Total duration (wall): 11m 7.2s
   Total code changes:    157 lines added, 5 lines removed
   Usage by model:
       unknown              0 input, 1.3m output, 0 cache read, 0 cache write (Est. 33 Premium requests)

Just burned through > 30% of my Copilot Pro Premium requests with a few prompts. I was using copilot-cli with mcp-playwright, with --allow-all-tools.

I noticed the issue when I suddenly got an unexpected " ✗ Quota exceeded. Upgrade to increase your limit: https://github.com/features/copilot/plans".

Hey folks -- attempting to reproduce on my end so we can investigate this.

For people who are affected by this: do you see that many premium requests being burned on the github.com Copilot usage page? I suspect this is just a client-side rendering bug and not actually spending from your premium request usage

it's not a UI issue, the data looks consistent between the CLI UI and github.com.

I started with ~70% this morning and it's 30% now.

the CLI UI shows "Remaining requests: 30%"

Github->Settings->Copilot Settings shows: "Premium requests: 70.1%"

This is happening to me, I just accidentally blew through ~200 premium requests in about an hour. I can watch the % just dropping percent by percent as claude sonnet 4.5 works on a single task.

Hey folks,

I'm reopening this issue until we've handled all billing issues / refunds / credit replenishments. This is still a work in progress and we will work to provide transparent updates as we make progress here.

The current state:

PRU billing overages (as they are reported to the backend) were fixed in .345
There is still a bug where incorrect PRUs consumption stats are displayed on the frontend, but rest assured we have made sure that the correct data is being recorded server side and reported for billing
There is a PR open to correct the client side display issue, and we will be cutting a release shortly with that fix
We are working internally to put together a plan to rectify any overbilling that may have occurred

Thank you for your patience as we work through this and we will keep you updated. In the meantime, please make sure you update to the latest CLI release.

@EvanBoyle

Hi folks, another update for you:

0.0.347 fixes the client side display issue explained by @EvanBoyle above.

We're still working internally to rectify overbilling! Thank you for your continued patience!

Wanted to check in here - I am using the latest version of the cli and I hit my premium request threshold at some point last week.

As far as I can tell my org has billing budget set up but I don't see my usage on the ORG level. I can only see the usage on my person copilot settings context:

The org copilot page shows that the premium billing is enabled:

The budget page shows a budget here:

Now, is the budget set up correctly? Not sure at all. Any advice appreciated.

Running any query results in the following:

 > test

 ✗ Model call failed: {"message":"You have no quota","code":"quota_exceeded"} (Request ID: 453B:14E4D4:711354:7EAD98:68FA4BC7)

 ✗ Quota exceeded. Upgrade to increase your limit: https://github.com/features/copilot/plans

Running /usage shows no info, but similar behavior exist in claude for the first message until it can handshake the API (I think).

 Total usage est:       0 Premium requests
 Total duration (API):  0.0s
 Total duration (wall): 2m 33.7s
 Total code changes:    0 lines added, 0 lines removed

version:

0.0.349
Commit: 3469b3e

debug output:

2025-10-23T15:36:33.938Z [DEBUG] InitiatorHeaderProcessor: Setting X-Initiator to 'user'
2025-10-23T15:36:33.938Z [START-GROUP] Sending request to the AI model
2025-10-23T15:36:34.439Z [DEBUG] Failed to get response from the AI model: CAPIError: 402 You have no quota
2025-10-23T15:36:34.439Z [DEBUG] Inner error: {
  "message": "You have no quota",
  "code": "quota_exceeded"
}
2025-10-23T15:36:34.439Z [END-GROUP]
2025-10-23T15:36:34.440Z [DEBUG] Command threw error: 402 You have no quota
CAPIError: 402 You have no quota
    at t.fromAPIError (file:///home/nonroot/.bun/install/global/node_modules/@github/copilot/index.js:1842:879)
    at HYe.getCompletionWithTools (file:///home/nonroot/.bun/install/global/node_modules/@github/copilot/index.js:1842:5268)
    at process.processTicksAndRejections (node:internal/process/task_queues:95:5)
    at async zce.getCompletionWithTools (file:///home/nonroot/.bun/install/global/node_modules/@github/copilot/index.js:1843:1334)
    at async t.getCompletionWithTools (file:///home/nonroot/.bun/install/global/node_modules/@github/copilot/index.js:2392:54798)
    at async wPn (file:///home/nonroot/.bun/install/global/node_modules/@github/copilot/index.js:2417:19001)
    at async Jul (file:///home/nonroot/.bun/install/global/node_modules/@github/copilot/index.js:2417:16739)

After running Copilot CLI workshops with over 100 participants, none reported any overbilling issues. I can confirm that both the Terminal/CLI usage display and the GitHub Copilot Settings page are now accurate.

Thanks for your prompt responses and support, Copilot team 🙏

I’ll leave it to you to close this issue when you feel ready.

I'm noticing my premium quota ticking down from 100% this morning as if it's been reset. Is this part of the remediation work?

The Enterprise portal still shows the same counts & overages currently.

same here... shows 100% too

Hi folks! We've fixed all the relevant issues here, and the billing team has reached out to affected users informing them of quota resets and overage refunds. Thanks again for your patience and help!

Just burned through 13% of quota within a single day. How can I "debug" this?

Navigation Menu

Repository navigation

Version 0.0.344 · Commit c87f222 - Possible bug in premium request counting (FIXED IN 0.0.345) #351

Describe the bug

Affected version

Steps to reproduce the behavior

Expected behavior

Additional context

13 remaining items

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Version 0.0.344 · Commit c87f222 - Possible bug in premium request counting (FIXED IN 0.0.345) #351

Description

Describe the bug

Affected version

Steps to reproduce the behavior

Expected behavior

Additional context

Activity

mattchengg commented on Oct 18, 2025

tat commented on Oct 18, 2025

0xHarlan commented on Oct 18, 2025

nathfavour commented on Oct 18, 2025

RyanHecht commented on Oct 18, 2025

ming86 commented on Oct 18, 2025

0xHarlan commented on Oct 18, 2025

mattchengg commented on Oct 18, 2025

iwangbowen commented on Oct 18, 2025

w-m commented on Oct 18, 2025

tat commented on Oct 18, 2025

akonyer commented on Oct 18, 2025

13 remaining items

EvanBoyle commented on Oct 20, 2025

RyanHecht commented on Oct 21, 2025

laz-jamesfuller commented on Oct 23, 2025

JakeLin commented on Oct 24, 2025

RDalziel commented on Oct 24, 2025

nathfavour commented on Oct 24, 2025

RyanHecht commented on Oct 28, 2025

deyanp commented on Dec 1, 2025

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions