OpenAI Core Product Head Tibo Sottiaux: Astra Expands to All Plus Users
Tibo Sottiaux announced that the company has begun rolling out GPT-6 Astra, aiming for a steady and rapid pace, and emphasized the need to cover all Plus users, not just Pro, business, and enterprise tiers. The rollout will take several days, with multiple new systems operating at scale for the first time and increased computing power. He also mentioned that all tiers will include Astra in their regular usage quotas, which can be fully utilized for Astra.
According to OpenAI's official website: Astra will first be available to a limited number of organizations on the same day, followed by a rollout to all Plus, Pro, business, and enterprise users, as well as interfaces with Amazon Cloud in the following days; usage will be included in existing subscription quotas, with additional points available for purchase. Pro, business, and enterprise users will also receive GPT-6 Astra Pro. The enterprise workspace is disabled by default and must be enabled by administrators. The website states that the Codex runtime framework will be upgraded to enhance computer usage speed, and when combined with Astra's efficiency, the Mind2Web task completion speed will be approximately 1.9 times that of the current GPT-5.6 Sol experience. Sottiaux currently oversees the core products of ChatGPT and Codex, having previously led the growth of programming agents and facilitated the merger of the two. Astra's release was delayed in August due to cybersecurity capabilities reaching critical framework levels, and advanced utilization capabilities are still undergoing limited testing.
Who is buying and who is selling: The event is driven by placing the flagship model into the 20 million Plus pool, rather than just selling to enterprise seats. Buyers are individual developers who have paid the $20 tier and want to fully utilize the new model, while sellers are laboratories switching and scaling computing power within their quotas. Beneficiaries include the unified entry point for Codex and ChatGPT, while those under pressure include usage-based interface clients and administrators who have not yet enabled the enterprise switch. Funds are flowing from Plus renewals and excess points into the newly scaled reasoning stack, while network capabilities remain separately gated.
Source: Public Information
ABAB AI Insight
Integrating Astra into Plus rather than locking it in Pro treats the "critical-level" model as a weapon for customer acquisition and retention. The enterprise is disabled by default, effectively leaving legal responsibilities to administrators while providing a trial experience for individual subscribers. The quota can be fully utilized for Astra, which will temporarily displace old model calls, exposing whether the new stack's computing power can truly rise as promised.
The capital path involves feeding the same model to both chat and programming agents, with Mind2Web acceleration targeting computer usage, as that is demonstrable professional work. Money flows from the predictable recurring revenue of the Plus pool, hedging against delays caused by critical-level security gates. Sottiaux's promotion from Codex head to core product lead indicates the company's bet on merging entry points rather than maintaining a separate product line solely for programmers.
Comparative objects include the tiered naming of GPT-5.6 (Sol, Terra, Luna) and Claude's tiered access to the strongest network and coding models. The industry position has shifted from research forecasts to expanding the subscription pool, relying on seats for growth and controlling through workspace switches and capability package splits.
The structural change is capital concentration. The mechanism is that once flagship intelligence enters the largest paying pool, price competition turns into quota competition; whoever allows regular subscribers to fully utilize the strongest model first will lock in developers' daily paths and then sell higher-tier Pro variants to enterprises for compliance.
ABAB News · Cognitive Laws
The flagship model must enter the largest paying pool to be considered truly released.
If the quota can fully utilize the new model, the old model will be pushed out of daily use.
Enterprise being disabled by default leaves the responsibility switch to administrators.