OpenAI reportedly launched GPT-6.1 Sol Ultrafast and improved Codex real-time course correction
BIBIBI
AT A GLANCE
The news flash said that OpenAI began rolling out GPT-6.1 Sol Ultrafast, and improved Codex real-time course correction during task execution.
Article
OpenAI 28 day sprint Day4: Sol Ultrafast's maximum 8 x speed,Codex supports real-time course correction
Beating AI news flash,OpenAI Product and Platform Lead Thibault Sottiaux announced 28-day product sprint's Day 4。GPT-6.1 Sol Ultrafast begins rolling out to API、Codex and ChatGPT Work. Officially, it has model capabilities close to GPT-6 Astra, and in Codex its generation speed reaches up to 8 times that of Sol Standard, and API up to 6 times.Ultrafast's API price is per million input token 12 USD, output token 60 USD, both 6 times the Standard tier.API is available to all developers, but subject to rate limits.Codex and Work are available only to Pro users paying 500 USD per month, as well as eligible enterprise and education users.
On the same 1 day,OpenAI also improved Codex's mid-task course-correction feature. In the ChatGPT desktop app,Agent while executing tasks, users can continue adding requirements or correcting the direction. The model can adjust faster, reducing cases where it continues working in the wrong direction.
Original link https://m.theblockbeats.info/flash/370899
Key points
01
OpenAI Product and Platform Lead Thibault Sottiaux announced the 28-day product sprint's Day 4。
02
The news flash said that GPT-6.1 Sol Ultrafast began rolling out to API、Codex and ChatGPT Work.
03
Officially,Ultrafast in Codex generation speed is up to 8 times that of Sol Standard, and in API it reaches up to 6 times.
04
Ultrafast API price is per million input token 12 USD, output token 60 USD, both 6 times the Standard tier.
05
API Available to all developers, but subject to rate limits.
06
Codex and ChatGPT Work are available only to Pro users paying 500 USD per month, as well as eligible enterprise and education users.
07
In the ChatGPT desktop app,Agent while executing tasks, users can continue adding requirements or correcting the direction.
AI-assisted interpretation
The following is analysis, separate from reported facts. Verify important claims independently.
This news flash introduced a faster but more expensive Sol tier, as well as Codex mid-task course-correction improvements. During Agent task execution, users can add requirements or correct the direction, and the model is said to adjust faster.
Why it matters to readers
The reported maximum speed, pricing multiples, and availability directly affect developers' usage costs and usability; mid-task course correction may reduce cases where Agent continues executing in the wrong direction.
Beginners should note that “maximum 8 x speed” describes only the peak value in specific scenarios;API also has rate limits, and Codex and Work eligibility is also restricted by plan and user type.
Risks and unknowns
The material is a retelling from the “Beating AI news flash” and does not include OpenAI an official announcement or pricing document; the specific products, prices, and rollout scope remain to be verified.
“model capabilities close to GPT-6 Astra” is not accompanied by a testing method or specific metrics.
API The specific rate-limit figures are not provided.
The criteria for “eligible” enterprise and education users are not provided.
The actual effectiveness and full rollout scope of the mid-task course-correction feature are not provided.
Related Developments
Loading event timeline…
Related concepts
API
This term is not in the glossary yet. Browse related concepts in the glossary.
