bug: spark usage limited when global usage limited
Resolved 💬 5 comments Opened Apr 27, 2026 by midischwarz12 Closed Apr 27, 2026
💡 Likely answer: A maintainer (github-actions[bot], contributor)
responded on this thread — see the highlighted reply below.
What issue are you seeing?
When there is GPT-5.3-Codex-Spark usage left and the Global weekly limit is reached, it is impossible to use GPT-5.3-Codex-Spark with the following error: You've hit your usage limit.
From /status:
│ Context window: 47% left (142K used / 258K) │
│ 5h limit: [████████████████████] 100% left (resets 18:25) │
│ Weekly limit: [░░░░░░░░░░░░░░░░░░░░] 0% left (resets 13:43 on 28 Apr) │
│ GPT-5.3-Codex-Spark limit: │
│ 5h limit: [████████████████████] 100% left (resets 19:30) │
│ Weekly limit: [████████████████████] 100% left (resets 14:30 on 4 May) │
│ premium limit: │
╰────────────────────────────────────────────────────────────────────────────────────────╯What steps can reproduce the bug?
- Hit global usage limit
- Switch model to GPT-5.3-Codex-Spark
- Send prompt
What is the expected behavior?
I should be able to continue using spark still when the general, global weekly usage window is fully consumed.
Additional information
_No response_
5 Comments
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
Credits are not relevant to this issue. This is about the usage limit through the plans.
Its currently not possible to switch an existing conversation from another model to spark when your main usage is limited. This is because spark has a different context limit, and we need to do compaction on an existing model using the gpt-5.3/5.4/5.5 model first, in order to reduce the context low enough that spark can use it.
<img width="833" height="261" alt="Image" src="https://github.com/user-attachments/assets/aa4ca416-5f65-4bbb-8257-0b8ff4ae341d" />I can reproduce the same issue.
Plan: ChatGPT Pro
Platform: Windows / JetBrains Rider
Standard Codex usage: exhausted
GPT-5.3-Codex-Spark usage: still available
After the standard usage limit was reached during automatic context
compaction, switching the existing thread to GPT-5.3-Codex-Spark did
not work.
The session repeatedly returned:
"Error running remote compact task: You've hit your usage limit"
Spark has a separate usage limit, but the remote compact task appears
to remain attached to the previously selected GPT-5.6 model.
Starting a completely new Spark thread may work, but the existing
conversation remains unusable. This prevents users from continuing
with their remaining Spark allowance.
Please ensure that compaction uses the currently selected model, or
allow the pending compact task to be cancelled when switching models.
@willwang-openai If I use an external ollama or llm.cpp model for compaction, I should still be able to switch to spark after. So your reasoning is flawed and does not fully consider the full feature set of codex.
To be fair, it took me until now to think of this as well.