Skip to content

OpenAI

Codex reset hunters lose Day 1 to a 50% speedup

4 min read

Written by an AI model from the linked sources and checked against them by a second model. How we work

All about: Coding agents →
Image: thenewstack.io

Thibault Sottiaux, engineering lead for Codex at OpenAI, said on October 5 that GPT-6 Astra and GPT-6.1 Sol now run about 50% faster by default. That is the first ship in a 28-day run in which his team owes a “clear improvement” every day or a full usage reset instead, The New Stack reports. Day 1 pays out nothing, and subscribers draining their allowance in the hope of a missed day are betting on a pledge that exists only in Sottiaux’s posts, with no word yet on which tiers a reset would cover or what it would restore.

In a follow-up post Sottiaux put the change at roughly 50 tokens per second, up from about 30. Taken literally, that is closer to a two-thirds gain than 50%, though both numbers look like rough estimates rather than benchmark results. Nothing needs switching on: subscribers who use Sign in with ChatGPT get the speedup in OpenCode, Pi, Amp and Devin as well as in OpenAI’s own products. “Let the improvements begin,” Sottiaux wrote.

What the bet is worth

Even the calendar is unofficial. Sottiaux set the challenge on October 4, and one tracker counts the run from October 5 through November 1 while noting that it is making its own call on the dates. A “full reset” can also mean very different things in Codex, which runs shorter rolling windows, such as the five-hour cap, alongside weekly limits. When you have run dry, getting one of those back is not the same as getting the other.

This has gone wrong before. After the April reset for “ALL paid plans,” a Plus user filed a GitHub issue saying their weekly limit stayed on its normal cycle, and Codex gave them no way to tell whether the special reset had applied at all. OpenAI Support later clarified that the reset did not cover every quota type for every user, the issue says.

Developers spotted the gamble within hours, and one tracker put up a 28-day scoreboard. One joked that Sottiaux had turned Codex usage into gambling: burn through the allowance every day and hope OpenAI misses a ship. Others would rather have higher limits to begin with. A Plus subscriber said the five-hour cap already dictates when they can work, and Mark Kretschmann, a software engineer at ASYS Group, read the pledge as 28 Codex resets in 28 days.

Resets as currency

Usage works as a stake because OpenAI has spent months handing out resets. When Sottiaux reset Codex rate limits for all paid plans in April, he opened by joking that resets cost money and shouldn’t be given out for fun, then conceded the vibes were good. In July, after the launch that folded Codex into the ChatGPT desktop app introduced regressions in some multi-agent workflows, he reset usage twice in one day, and another reset marked Codex and ChatGPT Work passing 8 million active users. At DevDay on September 29, he and Sam Altman gave out a banked reset live on stage.

GPT-6.1 Sol launched the same day. On October 1 Sottiaux posted: “Global reset landing tomorrow 10am PST for all paid ChatGPT accounts. Apologies for the slow start with GPT-6.1 Sol, it’s now back to running at expected speeds after the massive load spike in the first two days.” So the sprint’s first official ship speeds up a model whose slowness had already cost OpenAI a global reset that same week.

Not everyone is rooting for the reset button. The pseudonymous X account Token Gremlin argued that 28 smaller improvements won’t count for much unless OpenAI ships a model that can beat Claude Opus 5.5. T3 Code creator Theo Browne said on October 3 that Anthropic’s newest Opus had become the first model to pass 50% of traffic in T3 Code, with “literally half of all prompts” going to it.

October 30

Starting October 30, close to the end of the 28-day window, OpenAI cuts the Codex and ChatGPT Work allowance on the $200 Pro plan from 20 times the Plus allowance to 10 times. It paired that change with a new $500 plan at 25 times Plus and an Ultrafast tier for GPT-6 Astra that promises up to eight times the standard speed in Codex. Current limits hold through October 29. After that, existing subscribers get 62,500 usage credits, which OpenAI valued in September at $2,500 and which expire December 31, 2026.

No comments yet. Be the first.

Join the conversation

Sign in to leave a comment. It appears after moderation.