Issue: GPT-5.3-Codex-Spark appears to be blocked by the general Codex quota despite having a separate Spark meter.

What issue are you seeing?

Issue: GPT-5.3-Codex-Spark appears to be blocked by the general Codex quota despite having a separate Spark meter.

What I observe:

• My general Codex quota is exhausted.
• My GPT-5.3-Codex-Spark meter still shows available quota.
• However, when I select/use GPT-5.3-Codex-Spark, it is still blocked due to quota.

Why this is confusing:
OpenAI’s public/product wording suggests Spark has separate model-specific limits and should not count against standard Codex limits. If Spark still depends on the general quota being available, the UI should make that explicit.

Expected behavior:
Either:

1. Spark should remain usable while Spark quota remains, even if general Codex quota is exhausted; or
2. The UI should clearly state that Spark quota is separate but still requires general Codex quota/access to be available.

Impact:
The current meter display is misleading because it shows remaining Spark quota that cannot actually be used.

What steps can reproduce the bug?

Exhaust your general quota, leave Spark quota available, and then try to use Spark.

What is the expected behavior?

_No response_

Additional information

_No response_