Card #428 · OpenAI Platform & APIs
GPT-5.3-Codex-Spark context compaction and quota gating issues
Users are experiencing structural failures with GPT-5.3-Codex-Spark where context compaction causes the model to forget validated states and enter infinite loops. Additionally, there is a bug where the app blocks access to the Spark model even when its specific quota is still available.
Pain score
0.20/ 1.00 · weighted product of four components
Reach0.61distinct authors (log-normalized)
Recency1.00freshness decay, half-life months
Engagement1.00upvotes + comments, normalized
Monetization0.50willingness-to-pay cues in evidence
§ Evidence — 4 verified quotes
after a context compaction event, it appeared to lose track of the fact that those commands had already completed successfully.
the /goal state does not seem robustly survive context compaction.
the exhausted standard Codex weekly limit seems to hard-lock the whole composer and prevents Spark from being used as well.
Context automatically compacted followed immediately by: Your input exceeds the context window of this model.
Comments are a Pro feature — join the conversation, share what you've shipped, hear what others are building.
Upgrade to Pro