Codex ran out of room in the model's context window

Open 💬 4 comments Opened Feb 25, 2026 by Gusarich

thread id: 019c9584-a29e-7922-a030-4e015fd86bcc

happened with gpt-5.3-codex-spark, codex-cli 0.105.0-alpha.22, using exec

View original on GitHub ↗

4 Comments

Gusarich · 5 months ago

A few more sessions with the same problem:

019caff8-b07a-7541-ae6f-6f5b3c0c7707
019cafc6-9289-7a62-8f67-ceea16ec8857
019cafb2-cdbe-7c41-a667-6633c8d5ef4e
019caf87-3187-7992-b061-6cbde5ad5429
019caf82-d1d6-7011-9513-d71f01f76f22
019caf7c-53f4-75d0-a2d6-ab4631015dd4
019caf40-b50d-70f1-ab84-7b2f903195ec
019caf0a-5e55-7c33-8a6b-e30ef134e41e

I hope that helps!

Gusarich · 5 months ago

#13271 might be relevant

yassirusa · 4 months ago

When will this be worked on. Codex-5.3 and codex-5.3-spark are completely useless and they keep churning tokens without editing any code until they "run out of room"

emersonbusson · 3 months ago

This is the exact symptom tracked in #7808 — "Codex ran out of room in the model's context window" hit mid-turn is currently fatal to the thread. A proposed fix branch is linked from #7808 that recovers in place (mid-turn auto-compaction + one retry) rather than killing the turn.