Slow model response speed

GPT 6.1 Sol Medium has been so slow the last 48 hours or so. On tasks that used to take 2 minutes, it takes 10 minutes. And starting a new task always takes over a minute before it starts, saying it's Reconnecting. I can barely get through my usage because it's so slow.

@CliffCawley on X ↗ · 26 Oct 2026

Woke uo happy I used all my codex yesterday 🫡I hate how slow it is with these updates tho and Astra kinda laggy even in ultra

@QuillshiftGame on X ↗ · 26 Oct 2026

Day 1: GPT-6.1 Sol claims to have increased its speed by 50%; however, I have never seen a model this slow before. Are you sure any actual improvement has been made? This model is incredibly slooooow.

@telepati on X ↗ · 24 Oct 2026

It’s because the app is insanely slow

@AtaeiMe on X ↗ · 23 Oct 2026

downgraded main acct and canceled second acct, I literally can't burn all these tokens because the models are too slow and over-constrained

@technoascetic on X ↗ · 23 Oct 2026

Why is Codex getting slower and slower to respond now, with very slow output and frequent time out 5/5?

@ssshou110159 on X ↗ · 22 Oct 2026

@thsottiaux plz can you check this? Codex got very very slow. I use 6.1 sol and subagents, when start the thread is quite fast but its speed dramatically drops after around 30min executions

@IMEditor_ai on X ↗ · 21 Oct 2026

15 days in and I honestly have no clue what Tibo has actually shipped. Codex is still buggy as hell, usage limits feel like they got cut even more, and 6.1sol is still painfully slow. Had an issue Codex couldn't figure out after running for hours (literally a whole day), and…

@Mai97co on X ↗ · 20 Oct 2026

15 days in and I honestly have no clue what Tibo has actually shipped. Codex is still buggy as hell, usage limits feel like they got cut even more, and 6.1sol is still painfully slow. Had an issue Codex couldn't figure out after running for hours (literally a whole day), and…

@Mai97co on X ↗ · 20 Oct 2026

The reset thing is getting a bit tiring now... Claude's usage is in a different universe to Codex at the minute. Also feel like Astra has gotten 10x slower in past few weeks since this started.

@Aidan_Perston on X ↗ · 20 Oct 2026

Yep, sol6.1, running xtra slow as well

@VinishKal on X ↗ · 19 Oct 2026

What's the point of resetting usage limits when Codex is so painfully slow and unreliable that it's practically unusable? Higher limits mean absolutely nothing when the product barely works.

@skurovec on X ↗ · 15 Oct 2026

Today's Codex service stability is extremely poor, with continuous timeout reconnections. Even requests that do not time out have extremely slow response speeds, resulting in hours of wasted time. The Codex is nearly unusable; we need to fix it immediately. Thank you.

@FosenHai on X ↗ · 14 Oct 2026

What's the matter recently? Not only does it suggest that the model has insufficient capacity, but it is also super slow.

@MengFan786110 on X ↗ · 14 Oct 2026

Since this reset codex is painfully slow

@ronaldvdlee on X ↗ · 13 Oct 2026

Honestly, more usage isn't the fix. Your models just can't keep up with Claude right now. Astra burns through quota crazy fast and still isn't that strong, and 6.1 Sol is painfully slow, plus it way over-engineers everything. So double the limit or don't, we're not going to get…

@DavidsonDe55546 on X ↗ · 13 Oct 2026

But why is it so slow? I can’t do anything with this.

@johncalvo on X ↗ · 13 Oct 2026

The biggest problem is how slow it is—so slow that it's practically unusable! The spike in usage caused by the reset has only made the problem worse!

@Windsyu12138 on X ↗ · 13 Oct 2026

Honestly... I never though id say this but I think I might cancel my $100 GPT sub. Its just not worth it right now. Opus 5.5 is just too good and it feels unlimited no matter how much I use it and its out performing Astra (very slow and unusable) and of course Sol 6.1 (also slow)…

@jmaloneVS on X ↗ · 13 Oct 2026

This notification came way too late—I was enduring painfully slow 6.1 sol yesterday just to save on quota!!!

@Dranelie108 on X ↗ · 13 Oct 2026

我也提交申诉了，没啥用。。。响应速度也超级慢（as pro 500 sub)

@xiaomabosn on X ↗ · 13 Oct 2026

..........上下文提高了80%，速度降低了500%？ 今天怎么这么难用，跑了一天一个网页页面没跑出来

@Myskygood on X ↗ · 12 Oct 2026

I would like to use GPT at 60+ tokens/s at no additional cost; currently I only get under 30 tokens/s on a typical day, which makes heavy GPT-assisted office work painfully slow.

@qianrenyunjie1 on X ↗ · 12 Oct 2026

why is codex slow af now? its literally crawling. we need a new reset. i cant stand this anymore if this goes on i'm switching to claude max

@VCCnorthlon4evr on X ↗ · 12 Oct 2026

Performance got throttled again right after the reset. Please stop messing with GPT 6.1 Sol—I'm paying monthly, just leave me alone.

@DOANHDUNG0502 on X ↗ · 11 Oct 2026

Composer Predictions? How thoughtful! Codex is so damn slow these days that by the time it finishes last step, I've already forgotten what I was gonna do next. Glad you saw that coming. 😂

@KaranWen on X ↗ · 10 Oct 2026

reset!!!!! dots take so long i've had THE DOT compressing 24GLBS and its taken over 48 HOURS, pro 6 on web only takes like 2-3hrs at most

@ExposingC69997 on X ↗ · 10 Oct 2026

Cancelled. This feels completely out of touch with what subscribers actually want: speed, usage, and intelligence. Five days in, only Day 1 slightly improved speed, then it dropped again. The rest feels like distractions instead of fixing core problems.

@Kobbyhishmehzdu on X ↗ · 10 Oct 2026

Cancelled. This feels completely out of touch with what subscribers actually want: speed, usage, and intelligence. Five days in, only Day 1 slightly improved speed, then it dropped again. The rest feels like distractions instead of fixing core problems.

@Kobbyhishmehzdu on X ↗ · 10 Oct 2026

GPT is too slow right now, that's the primary problem that needs to be solved.

@zyf97060311 on X ↗ · 10 Oct 2026

Selected model is at capacity and very very very slow

@ustcguoyi on X ↗ · 10 Oct 2026

Pretty bad, considering the only decent upgrade was increasing the speed of Sol. And obviously you made it slower not faster, so the one upgrade was just a lie. So in reality you've give out a banked reset and a forced reset that messed up a load of people's budgeting. 2/10.

@danny_______ on X ↗ · 9 Oct 2026

Codex is too slow

@hahagewhahagewh on X ↗ · 9 Oct 2026

Actually, it's slower than before.

@sylvenPR on X ↗ · 9 Oct 2026

超级慢

@David217820 on X ↗ · 9 Oct 2026

How about you just fix the current product. About to get opus for a second opinion as codex and gpt just getting dumber by the day. Just repeating instructions over and over now. Usage is now better but takes forever and still doesnt produce the results.

@chrisrismay on X ↗ · 8 Oct 2026

super slow, prompt :Hi took 18 seconds

@KhihortM on X ↗ · 8 Oct 2026

Tibo, adoro codex, pero está inusable. Está extremadamente lento... al punto que no tiene sentido... 30 minutos para delegar una auditoría... la gobernanza que tengo está muy bien. no es el problema.

@gerarguello on X ↗ · 8 Oct 2026

the codex is fucking slow, what the hell have you done?????? dont let me subscribe claude

@sapphire_157_ on X ↗ · 8 Oct 2026

My codex running on 6.1 Sol Ultra is so slow.......Anyone got the same problem?

@William16816888 on X ↗ · 8 Oct 2026

我只想gpt6.1 sol不要这么慢，忍受不了了

@yuan_aigc on X ↗ · 8 Oct 2026

@thsottiaux codex 速度太慢了 太慢太慢

@YouquanJin91366 on X ↗ · 8 Oct 2026

@thsottiaux Models are so Slow, I can‘t use it for Everyday work atm. It‘s more like, Plan a Task and Go sleep. MAYBE After that it‘s ready.

@linklootio on X ↗ · 8 Oct 2026

GPT 6.1 sol minimal still extremly slow when it comes to powerpoint presentations

@Mohsine_Mahzi on X ↗ · 7 Oct 2026

am I the only one who is having speed issues in Codex?

@SpaceFacts_us on X ↗ · 7 Oct 2026

Tibo, My personal thoughts after updates: 1) Usage has decreased, but I believe the main reason is that the 6.1 Sol model runs too slowly, regardless of the inference intensity setting. 2) Other competitors to Codex have gradually opened up access to users in other countries.

@Kelvinplchan on X ↗ · 7 Oct 2026

not trying to be that guy but i’ve been using codex sol 6.1 all day yesterday and today, and it actually feels significantly slower today than it did yesterday pre-efficiency improvements. i’ve also updated to the newest release version. anyone else experiencing this?

@whiteteebhf on X ↗ · 7 Oct 2026

The standard speed of 6.1sol is really surprisingly slow! Long tasks are always stuck in one step, and I don't know what to do after consuming 1% of the quota.i want to chose claude！

@yixun_chen56878 on X ↗ · 7 Oct 2026

ran exactly that. same repo, same prompt, high effort, whole task wall clock sol 79s → 76s, astra 44s → 76s. all passed 12/12 claude control slowed too so some of it is load, rerunning off-peak tonight

@SergeForge on X ↗ · 7 Oct 2026

Hey @thsottiaux could you please check on your end this?. at the beginning yes the speed feels so good. but since yesterday and today is feeling pretty slow again

@xvierdavid on X ↗ · 7 Oct 2026

it was pretty weak today, but I would say reset tomorrow. Day 1 was good in theory but actually not that great, speeds on SOL 6.1 still slow.

@voicesintime on X ↗ · 7 Oct 2026

Neither, codex is too slow for me to do any productive work. Dot is not here, usage cut in half, Maybe the ultrafast on sol 6.1 might help me use the resets before plan cancels, can we ship that?🥸

@M_Wahib01 on X ↗ · 6 Oct 2026

What we are after is very snappy and fast output, like what we see in Opus 5.5, but with the reliability and veracity of Astra/Sol. Actually, 50 is not working at all—it just keeps running without any indication of speed, so no, we do not see any visible change. We also need…

@madanpandit on X ↗ · 6 Oct 2026

Wait, why did this blow up with likes?)) Don't get me wrong, ChatGPT/Codex is a great product. But the speed is genuinely painful compared to competitors, and that $500 plan is crazy. Zero beef with Tibo personally. My only point, and I really want to stress this, is that you…

@d_ridelman on X ↗ · 6 Oct 2026

Sorry, but is still using the same output speed on my end And yes I used Sol 6.1 on Ultra with x1.5 faster results, but inference speed keep looking bad also there is time where the model starts doing nothing and your limit disappears without the model just doing anything.

@MikeGonzalezDEV on X ↗ · 6 Oct 2026

I dont see any improvements on speed. GPT 6.1 Sol is suppppppeeeer slow. You can wait from 30 to 60 min for a simple updates. it's hilarious when you try claude opus 5.5 it's on the other level. What about you guys?

@paw_pixels on X ↗ · 6 Oct 2026

what do u mean by default speed? it seems very slow like before

@ijaack94 on X ↗ · 6 Oct 2026

Codex tasks assigned and managed by my dot in the cloud, with execution on my own computer, are still extremely slow. The context seems to fill up and get compacted repeatedly, and I suspect this may be contributing to the delays. I haven’t confirmed the cause, but I’m not seeing…

@pcyuan30 on X ↗ · 6 Oct 2026

restest complete, 3 of 4 requests back inside launch day speeds. thanks @thsottiaux

@BrenBuilds on X ↗ · 5 Oct 2026