Sudden increase in GitHub Copilot Premium requests

[yrademordeman]

Question

Did anything change recently in how OpenCode usage is counted toward GitHub Copilot Premium requests?

Noticed a big jump in GitHub Copilot Premium requests last few days with no change in my usage patterns 😅

[github-actions]

This issue might be a duplicate of existing issues. Please check:

• Copilot auth now sets far too many requests as "user" consuming premium requests rapidly #8030: Copilot auth now sets far too many requests as 'user' consuming premium requests rapidly
• Github Copilot multiple premium request charges without subagents #8067: Github Copilot multiple premium request charges without subagents
• GitHub copilot premium calculation #1175: GitHub copilot premium calculation

Feel free to ignore if none of these address your specific case.

[anasmohammed361]

i noticed it too , but not sure why this happened

[adrians5j]

+1

Closely watching for changes related to this msg.

[yrademordeman]

Ah, thanks! Completely missed those issues.

So to summarize.

It's partially due to properly counting premium requests i.e. 1 / message rather than 1 / session.
But potentially also something else...

[rekram1-node]

You can watch this:
#8393

Waiting on copilot team before we can release it.

But yeah youll notice increased spend because it was counting requests incorrectly before

[dhruvkej9]

Ah, thanks! Completely missed those issues.

So to summarize.

It's partially due to properly counting premium requests i.e. 1 / message rather than 1 / session. But potentially also something else...

1 per message seems too much, now we can't give AI small followups which we forgot else it would cost us too much, in case of Opus 4.5 it will burn the pocket

[yrademordeman]

Agreed, I will definitely have to change my workflow or find another provider. I burnt trough all of my premium requests in 2 days before noticing. Though, I guess that was how it was supposed to work all along.

I'll probably try using opus 4.5 through Antigravity subscription next. Not sure how bad the rate limits are.

[adrians5j]

Wait a minute... so you guys are saying this new sudden increase in premium requests is how it should've worked already?

In other words, until 1.1.13, tokens did not get counted correctly and now that is the case?

I feel I'm missing something. 🤷🏻

IDK, maybe I just need to give it a quick test. I'll do that and see how it behaves. I'll be using Webstorm's Copilot plugin and OC and see what are the diffs in terms of token consumption.

[dhruvkej9]

Wait a minute... so you guys are saying this new sudden increase in premium requests is how it should've worked already?

In other words, until 1.1.13, tokens did not get counted correctly and now that is the case?

I feel I'm missing something. 🤷🏻

IDK, maybe I just need to give it a quick test. I'll do that and see how it behaves. I'll be using Webstorm's Copilot plugin and OC and see what are the diffs in terms of token consumption.

It was working partially correctly but now it is worse, earlier follow-ups didn't cost any premium while the agent is working, but now it counts.

Earlier each user message counted premium not the way they are saying that only one request per session, you can test 1.1.13 and sonnet 4.5 it costs 1 premium request for each user message everytime (not for follow-ups).

Maybe the problem was for some models but some were working fine.

[yrademordeman]

From my understanding this is how using copilot in any other tool works (github webui, copilot cli, copilot in vscode etc.)

See e.g. https://docs.github.com/en/copilot/concepts/billing/copilot-requests#what-are-premium-requests

• Copilot Chat uses one premium request per user prompt, multiplied by the model's rate. This includes ask, edit, agent, and plan modes in Copilot Chat in an IDE.
• Each prompt to Copilot CLI uses one premium request with the default model. For other models, this is multiplied by the model's rate.
• Each time Copilot reviews a pull request (when assigned as a reviewer) or reviews code in your IDE, one premium request is consumed.

...and so on.

There might have been a bug in opencode that caused it to count some tool calls as user messages which caused it to over count premium requests a bit. I think that was fixed when they added the official github copilot plugin https://github.com/anomalyco/opencode/releases/tag/v1.1.21

However I don't think 1 premium requests / session is coming back 😢

[dhruvkej9]

From my understanding this is how using copilot in any other tool works (github webui, copilot cli, copilot in vscode etc.)

See e.g. https://docs.github.com/en/copilot/concepts/billing/copilot-requests#what-are-premium-requests

• Copilot Chat uses one premium request per user prompt, multiplied by the model's rate. This includes ask, edit, agent, and plan modes in Copilot Chat in an IDE.
• Each prompt to Copilot CLI uses one premium request with the default model. For other models, this is multiplied by the model's rate.
• Each time Copilot reviews a pull request (when assigned as a reviewer) or reviews code in your IDE, one premium request is consumed.

...and so on.

There might have been a bug in opencode that caused it to count some tool calls as user messages which caused it to over count premium requests a bit. I think that was fixed when they added the official github copilot plugin https://github.com/anomalyco/opencode/releases/tag/v1.1.21

However I don't think 1 premium requests / session is coming back 😢

You can downgrade your version and use it

[yrademordeman]

You can downgrade your version and use it

I'm guessing that's against githubs terms of service and you risk getting your account suspended. Looks like that already happened to some people previously.