GPT-5.5 Fast spent 12+ minutes thinking, produced no output, then entered reconnecting state

Codex app version

Desktop Codex: 26.608.12217 — released Jun 8, 2026

Platform / plan

• macOS Tahoe 26.5.1
• Apple Silicon MacBook Pro
• Pro plan
• Mode: GPT-5.5 Fast
• Reasoning: Extra High

Related feedback ID

019eb48b-00d0-7b91-b50a-cbd119bb5c4c

Summary

I encountered another serious GPT-5.5 Fast performance issue.

The model spent over 12 minutes in the thinking state without producing a single token of visible output. After waiting for more than 12 minutes, the task entered a reconnecting state instead of continuing normally.

Because there was still no useful output and the session appeared stuck, I was ultimately forced to stop the task manually.

Observed behavior

• GPT-5.5 Fast started processing normally.
• The task remained in a thinking state for approximately 12 minutes.
• No visible response content was generated during that period.
• The session then entered:

Reconnecting 3/5

• After reconnecting appeared, there was still no meaningful progress.
• I eventually stopped the task manually.

Expected behavior

If GPT-5.5 Fast needs additional reasoning time:

• it should gradually stream progress or output,
• it should not remain silent for 12+ minutes,
• and it should not fall into reconnecting before producing any useful result.

At minimum, users should receive either:

• partial output,
• progress updates,
• or a clear explanation of what the model is doing.

Actual behavior

The model spent more than 12 minutes thinking, generated zero visible output, then entered reconnecting and became effectively unusable.

Why this matters

This behavior is extremely disruptive because users cannot tell whether:

• the model is genuinely working,
• the task is stalled,
• the backend is overloaded,
• or the session has already failed.

In practice, it causes significant time loss because users may spend more than 10 minutes waiting only to discover that no progress was made at all.

Request

Please investigate why GPT-5.5 Fast can remain in a thinking state for more than 12 minutes without generating output and then fall into reconnecting.

The model should either:

• produce incremental output,
• provide meaningful progress feedback,
• or fail fast with a clear error message instead of leaving users waiting for extended periods with no visible progress.

<img width="1782" height="640" alt="Image" src="https://github.com/user-attachments/assets/0f6e0c26-973b-46c5-8657-5e528a719b33" />

<img width="1606" height="556" alt="Image" src="https://github.com/user-attachments/assets/bdd5025d-cbc0-4fa2-9172-285c5a4c3cca" />