Thibault Sottiaux, the OpenAI product lead known as Tibo and responsible for Codex and ChatGPT, said on Oct. 6 that the default speed of GPT-6 Astra and GPT-6.1 Sol for subscribers had increased by about 50%, rising from 30 tokens per second to 50 tokens per second. He said the change also applied across OpenAI products and third-party tools connected through Sign in with ChatGPT, and that users would not need to make any changes on their side. The effect, he wrote, should be noticeable within roughly two hours.
In his post, Tibo wrote: 「Day 1 / We have optimized the default speed to be ~50% faster across GPT-6 Astra and GPT-6.1 Sol through the subscription across all our products and partners using Sign in With ChatGPT... No changes needed on your end and this should be...」
Day 1 focused on speed
The update was presented as the first delivery under a 28-day pledge Tibo made on Oct. 4 U.S. time. He said that over the next 28 days, the team would either ship one improvement each day that most Codex and ChatGPT Work users could clearly feel, or fully reset usage quotas.
Speed was the first item because GPT-6.1 Sol had run into problems shortly after launch. The model went live on Sept. 29 and slowed during its first two days as demand surged. Tibo said it was close to being the highest-demand model in OpenAI’s history, that the team had added capacity, and that speed could get close to twice the previous day’s level. He later apologized for the “slow start” and said all paid accounts would be reset.
That puts the Day 1 delivery in direct context: the first fix targeted the weakness he had already apologized for.
Four priorities had already been laid out
In the post before making the 28-day pledge, Tibo said the team would focus on four areas next: simplification, getting more usage through efficiency, breakthrough features, and new models. He said user feedback had been clear and that people wanted simpler products.
Still, his speed-update post did not mention any changes to usage caps. It also did not include a measurement method or any third-party benchmark. For now, the 50 TPS figure remains a number provided by Tibo himself.
Day 2 brought four more updates
On Day 2, which the report identified as Oct. 7 Taiwan time, Tibo announced four more changes:
- Auto-review became free and no longer counts against quota.
- API paid usage tiers were reduced from five levels to three.
- A meeting notes plugin was introduced.
- The Decisions API entered public beta.
Tibo described Auto-review as a second agent reviewing the main agent’s actions to block high-risk operations. He said enabling the feature could previously consume 2% to 10% of a plan’s quota, but that usage would no longer be counted.
Poll results showed most users still wanted a reset
After posting the Day 2 updates, Tibo also ran a poll asking users how the day felt. The poll drew 74,565 votes. Of those, 76% selected 「needs a reset」 and 24% chose 「good day」.
As of 10 a.m. mentioned in the report, the reset still did not appear to have happened.