@OpenAI_Support this issue is shown itself again today. It started around 15Z JUNE23.
Also a lot of hallucinating today.
Please advise.
Using Model 5.5
@OpenAI_Support I realized that my toggle is using model 5.5 but after the prompt is generated it says it utilized 5.3 - see additional commentary below.
I want to surface something that may be useful to others working heavily in structured writing, strategy, or AI-assisted workflows.
Over the last several sessions, I’ve noticed inconsistencies in model routing behavior while working in ChatGPT projects — specifically where a selected higher-capability model appears to fall back to a “mini” runtime identifier in system outputs.
This isn’t about output quality in a single instance — it’s about consistency in deterministic work.
When you’re building:
- executive-level documents
- capital allocation narratives
- structured operating models
- or high-stakes positioning documents
even small shifts in model behavior can introduce drift in tone, structure, and precision.
What I’m trying to understand is:
- Is this expected fallback routing under load?
- Is it a project-level orchestration behavior?
- Or is there a mismatch between selected model and runtime execution?
If anyone else working in enterprise, product, or AI workflows has seen similar behavior, I’d be interested in how you’re handling consistency across sessions.