I’ve been experiencing extremely slow response times and code generation with OpenAI Codex lately. Sometimes, it takes way too long to respond or complete coding tasks, making the whole development workflow frustrating.
The weird thing is that I’ve heard similar complaints from people using different subscription plans, including Pro. So I’m wondering whether this is a general issue with Codex right now or if it’s just happening to certain users.
A few questions:
- Are you experiencing the same slowdown with Codex?
- Does it happen on both Free and paid plans, including Plus and Pro?
- Is the slowdown affecting code generation, reasoning, or even simple responses?
- Has performance gotten worse recently, or has it always been this slow for you?
I’d appreciate hearing about your experiences, especially from people using Codex for serious, day-to-day development work.
Is this happening to everyone, or am I just unlucky?
Hey @mowolf I am Pro 200 User and No, there have been slight issues like “Model at its capacity” however I hadnt had any such experiences which make me determine that the response / reasoning ability or coding capability has been downgraded.
Yes to this x1000. I have a 285k, 4090, and 8400 cudimm. Not that it matters because codex doesn’t even know how to use any of it. That’s the main issue. It only uses 1 thread and can’t use the VRAM properly or if at all. It basically will cap itself out at 50% ram usage.
BUT the kicker is that it keeps itself open in the background. So what i realized i had to do it literally just end the process after every big task and re open. It’s just leaking your RAM.
I’m experiencing severe slowness too. I’m on the $200 Pro plan, and over the past two days the Codex desktop app has been so slow that I haven’t been able to get productive work done.
This is frustrating and, honestly, quite sad. I keep finding myself considering cancelling Pro and moving to Claude because I need to get my work done. For now, I’m trying to complete my tasks in Antigravity with Sonnet 5.5, where I also have a Pro subscription.
I’ve found other recent reports of Codex slowdowns, including a Windows desktop report filed on October 12 describing very long waits and stalled tasks. Experiences vary, including in this discussion, so I can’t assume everyone is affected or that these reports share the same cause.
Today’s OpenAI status update acknowledges degraded performance in GPT-6.1 Sol Codex and Work mode. It says a mitigation has been applied and recovery is being monitored. I appreciate that update, but I don’t know whether this incident explains everything I’ve experienced over the past two days.
Could the OpenAI team clarify whether the incident covers these desktop symptoms, what affected users should do if the slowdown continues, and what information would help investigate individual sessions? A clearer update on recovery would help me plan my work and decide whether to keep the subscription.
I want to be able to rely on Codex for my work again.
Yes, it’s very slow, the response times are slow, the code writing process, troubleshooting, reading code, everything is slow. I have the GPT-200 subscription, but I already regret it. Luckily I have the Claude Max 20x subscription; I jumped on the GPT bandwagon to use the Astra for gaming and I messed up.
I am a £200+ VAT User in UK.
Today was a nightmare for using Codex and so was yesterday.
Super slow response, constantly reconnecting, Model at capacity, you name it. It is so slow that it is almost not usable for power users in my opinion. Hope they sort this out right away and instantly.