At 17:20 UTC on 5 October 2026, Tibo (@thsottiaux) posted a note labeled Day 1. His X bio reads “Codex & ChatGPT @OpenAI”. The post says OpenAI has “optimized the default speed to be ~50% faster across GPT-6 Astra and GPT-6.1 Sol through the subscription” on OpenAI products and on partners using Sign in with ChatGPT, “including OpenCode, Pi, Amp, Devin, …”. It says no changes are needed on the user’s side, and that this “should be felt within the next two hours.”
The note quotes his 4 October post: over the next 28 days, each day the team will “either ship one thing that is a clear improvement and relevant for most codex/work users or ship a full reset.”
The OpenAI Developer Community thread restates that Day 1 item. Its summary says Astra and Sol “should now feel around 50% faster by default” through a subscription, names OpenCode, Pi, Amp, and Devin, and says the improvement should be noticeable within the next two hours. That summary closes the partner list at Devin. The post itself leaves an ellipsis after Devin.
Whose number it is
The ~50% figure is OpenAI’s, written with a tilde. The Day 1 post states the percentage and does not include a measurement method.
RuntimeWire attributes a second pair of rates to Sottiaux (it uses the name Thibault Sottiaux): roughly 50 tokens per second, up from 30, and describes those rates as approximate. The Day 1 post does not contain a tokens-per-second figure. RuntimeWire calculates that a move from 30 to 50 tokens per second is about a 67% increase if both numbers are exact. The Day 1 post says ~50%. Any tokens-per-second rate in this story is that secondary, vendor-reported approximation.
The Day 1 post names the subscription path. It does not mention the API. A community reply asks whether API speed changed. The thread contains no OpenAI reply to that question.
The thread quotes Vaibhav (VB) Srivastav calling the work an inference optimization, at “50% faster,” across Sign in with ChatGPT and plans, rolling out over the next couple of hours. Tibo’s post says “default speed” and “within the next two hours.” It does not use the word inference.
One reply attached the author’s own task timings and said the author was unconvinced the speed-up was helping GPT-6.1 Sol. That table is the reply author’s run.
What to do with it
On a ChatGPT subscription, including a partner app on Sign in with ChatGPT, OpenAI says the faster default is already in effect and that nothing has to be switched on. The claim is day 1 of a stated 28-day series for most Codex and Work users. On that series, a day without one clear improvement is described as a full reset. Treat the percentage as OpenAI’s approximate figure.
The Campfire
No commentsNobody has pulled up a log by this one yet. Be the first to say what you make of it.
Held for the desk. It appears after a look.