Codex 28, day 1 record

The day’s change

Recorded update

Improvement

Tibo announces roughly 50% faster default speeds

Tibo says default speeds for GPT-6 Astra and GPT-6.1 Sol have been increased by roughly 50% for subscriptions, across OpenAI products and partners using Sign in with ChatGPT, including OpenCode, Pi, Amp and Devin.

Read the whole note

Announced on 5 October at 17:20 UTC, with no user action required and rollout expected within two hours (by approximately 19:20 UTC). This is the announced performance improvement; Rantr has not independently benchmarked it. Try it on your own tasks before judging the result.

In a reply to the Day 1 announcement, Tibo adds that speeds are reaching 50 tokens per second instead of 30 and credits the tokenizer and token efficiency. These are Tibo’s figures, not a Rantr benchmark. Source ↗

In a later reply about Astra performance, Tibo says the Astra changes were rolled out earlier and that 50 tokens per second is the expected figure. This clarifies the rollout described above; it is not a separate release or proof of the speed in every account. Rantr has not independently verified it. Source ↗

Source ↗

6 useful · 5 not enough · 1 not tried

Votes cast by the close of the day. Later votes count on the campaign page, not here.

Running totals

What changed since the campaign began

Totals since 5 Oct 2026, at the close of each day; today’s are so far. A total never goes down when a sample is recounted.

experiences0no change
“Same here”0no change
update votes12+12
release notes0no change
release check answers, on any note0no change
replies collected34+34
corrections0no change

Replies on X

Replies to Tibo’s post published this day

A sample collected from X, not every reply. Each link opens the reply on X.

Replies dated this day in the collected sample: 34. Collected since 5 Oct 2026: 34.

X replies are found by search and each one is checked against its public post on X. Last read: 26 Oct 2026, 23:46 UTC. Not every reply is read.

An idea
Tibo, pls add a feature, where you can set your goal, and it'll autoresume it when your limit resets, or you can schedule that. That's like the only spot where claude has you beat, to resume my codex I have to set alarms or do it as soon as I get up. Pls add the goal auto resume.

@Aaron1559386 on X ↗ · 00:44 UTC
An idea
Feature #1 restore the $200 plan

@gmathis1995 on X ↗ · 00:52 UTC
An idea
From "your X community" 1. Remove the 5-hour usage limit on Plus for Codex and Work. 2. Make the Sol budget last a real session, not a single routine task. A 5-hour window should not die in half an hour of normal repo work. 3. Fix GPT-6.1 Sol speed. ~20 t/s on the standard plan…

@Rolandofj on X ↗ · 04:00 UTC
A problem
Don’t know about improvement, but apparently today saying "hi" to GPT-6 Luna Light in a brand new chat costs 1% of my weekly limit and 4% of my 5-hour quota.

@LooteraFriendly on X ↗ · 05:31 UTC
A problem
The resets are useless. They are useful only if you don't move the next reset date. So this challenge is self serving.

@shantanugoel on X ↗ · 06:09 UTC
A problem
I had a reset that expired on October 4th but it's 11:46 PM on October 4th still, we need a time of exactly when the banked resets expire because it was still available at 9:30 pm earlier tonight and I was just about to use it and now it's gone...and it's still October 4th

@chadfeingoldseo on X ↗ · 06:49 UTC
A problem
When using remote Codex via iPhone 13 ChatGPT app (Codex running on MBP M4), the "Send" button will show loading animation indefinitely after sending the first message and the agent has finished it. An easy fix to save 1 day of reset;)

@mhdmts on X ↗ · 08:11 UTC
A problem
I need to say OpenAI resets became totally useless. As they move the standard reset date. I’ve used my reset from Anthropic on 200 plan today, and the planned reset date is the same. So I plan several big tasks in the next 2 days - this is the real limit bonus

@VBandura8 on X ↗ · 10:34 UTC
A win
I made the switch last month to Opus, mainly for usage. I’d run out of weekly usage in a few days. But now I won’t go back. Opus just runs and runs. GPT will tell you it’s running, and you check in 15 minutes later and its like, sorry I never started.

@bridgerloftin on X ↗ · 13:19 UTC
A problem
When I change the model to Astra it cost more usage than before please find the problem and fix it

@Johnsmith0wbv on X ↗ · 16:17 UTC
A problem
Hey @thsottiaux, if your using Dot voice chat on your laptop and forget to close the session you can’t start a new one on the phone. Asking Dot to stop the voice session through text chat doesn’t work either. I’m locked out of voice until I get back to my laptop.

@scottmillersb on X ↗ · 16:32 UTC
Mixed
Literally, Chatgpt dot gave me the best usage of my life. I ran out of 100$ codex usage within 10 hours of the last reset. But ive shipped 3 apps using dot alone. By the way I had to do testing by myself since I had no codex credit for the hands on testing and verification.

@Nonosam2 on X ↗ · 16:41 UTC
An idea
Hey Tibo, can you make reset tokens auto-use just before they expire? It's been several times where I don't trigger it on the last day and I miss out on it, I think users would love you more if you auto applied after hitting zero (usage %) or if it's about to expire

@cocopuffs239 on X ↗ · 16:59 UTC
An idea
I had a banked reset set to expire on 10/05, but no specific time was provided. I assumed it would remain available until the end of the day, but now it's already gone. Really disappointing! Please add an exact expiration time so we know precisely when a banked reset will expire.

@g_steglich on X ↗ · 17:00 UTC
A problem
Ngl Tibo, I lost faith in you guys. Claude had outclassed you all in every aspect and I really wanted you guys to win. Computer use is really the only place you shine and the usage is so bad and slow that even that is a stretch. Cancelled my extra 100 sub. May cancel my 200 too.

@urmomsgpt on X ↗ · 17:02 UTC
A problem
spent over 80% of my usage correcting the errors that 6.1 sol keeps making T_T wished there was a reset today (or at least a birthday reset) hahaha

@Paul_Zachdev on X ↗ · 17:07 UTC
A problem
Same thing here, the reason is obvious: They ran out of VMs, and now they are crashing existing sessions left and right to get new users started.

@Ingenitu on X ↗ · 17:26 UTC
A problem
cloud threads stopped working for me. Please investigate.

@imjvdn on X ↗ · 17:33 UTC
A problem
I still hate the 5 hour limit in Plus…

@isIcedudu on X ↗ · 17:36 UTC
Mixed
okey, a bit faster not quite 50 but yes 😊

@showoffdennis on X ↗ · 18:31 UTC
A problem
Def could use some resets. The 10x plan ends up doing about 30-40 hours of work on only one of my projects before running out.

@workwithdavew on X ↗ · 18:48 UTC
A problem
Why is tokens usage becoming soo high? I finished my weekly limit in 1d although didn't increase usage ..

@wassimbs2 on X ↗ · 19:01 UTC
A problem
Is Codex Pro 20x usage accounting broken? Oct 4: 224M tokens used 10% of weekly allowance; Oct 5: 201M used 26%. Remaining allowance fell from 70% to 21%, including 30%→21% in under an hour. Separate bug: switching to Sol Light left existing subagents on Astra low.

@Ryhrw4y on X ↗ · 19:06 UTC
An idea
i need a codex reset on the $500/mo plan

@needcodexreset on X ↗ · 19:20 UTC
An idea
Please give Plus subscribers back the full access to their weekly limit. No 5-hour limit!

@sstaver on X ↗ · 20:08 UTC
A win
I canceled my subscription and switched to Claude Code, and suddenly things started moving. Funny timing.

@tomashined on X ↗ · 20:11 UTC
A problem
Tibo, I had a banked reset saved for today. The reset didn't say what time it was ending, and already at 1pm EST (just 10am PST) it was already gone from my account before I had a chance to use it. This is an awful, anti-consumer system.

@grubbygrubsters on X ↗ · 20:21 UTC
Mixed
restest complete, 3 of 4 requests back inside launch day speeds. thanks @thsottiaux

@BrenBuilds on X ↗ · 20:48 UTC
Mixed
Today I’m being a bit bold and going back and forth, asking Opus and then asking Astra. I’m nervous about using up a lot of the rate limit, but I thought this was the most efficient way.

@41u_jr on X ↗ · 22:11 UTC
A problem
PiでSign in With ChatGPTを利用していると、リセットしても利用上限エラーが発生して使えません。改修していただきますようお願いします。

@yukichi_AIB on X ↗ · 23:43 UTC
Mixed
I can definitely notice the speed increase, which is fantastic. However, my $200 limit still doesn't feel like enough. Could you please increase it a bit?

@Samuelhs on X ↗ · 23:57 UTC
Mixed
The speed is optimized, I can feel it but the following of instructions feels off now. Sol and Astra feel like they do no longer follow Agents.md as much as before. Is it just me?

@Rhaps0dy_Sky on X ↗ · 23:58 UTC
Mixed
GPT-6.1 Sol feels smarter and faster, but it burns through tokens a bit quicker than Astra.

@www_skstu_com on X ↗ · 23:58 UTC

1 X reply from this day is not quoted: Rantr does not have its words or held them back after automated screening. It stays on X.

Nothing this day for: OpenAI release notes, published experiences, corrections.

Method

How this record is made

Each day runs from 00:00 to 24:00 UTC. The record counts votes, release check answers, experiences, “Same here” and replies timestamped inside the day. When the day closes the record is frozen; it is recomputed only after an entry in the public corrections log, or when the day’s update is recorded late. Rantr is a neutral record: it does not take a side.

Corrections log → Open data and method →